RE: howto make system calls (newbie question)

RE: howto make system calls (newbie question)

[Beard, Frank]

I totally agree with Marin.
Frank

-----Original Message-----
From: Marin David Condic [mailto:marin.condic.auntie.spam@pacemicro.com]
Sent: Friday, May 04, 2001 11:49 AM
To: comp.lang.ada@ada.eu.org
Subject: Re: howto make system calls (newbie question)

While I'm basically on your side here, Marc, I think it is fair to point out
that in Kernighan & Ritchie (I don't have the ANSI C standard in front of
me) it says there is a "Standard Library" that, while "not part of the C
language proper" is considered to be part of the C environment that a
standard implementation of C will provide. In that library, we find
<stdlib.h> which contains a function "system" that one can reasonably expect
to be part of a standard C implementation.

I would consider this analogous to the Ada appendices that define things
like Unbounded_String, etc. The various string packages are not part of the
Ada language per se, but are an expected library of things available in Ada
if the implementation is conformant with the standard. I see no reason that
Ada couldn't provide similar libraries (perhaps optional for implementations
where it makes no sense to have them - e.g. embedded machines) for functions
like "system" so that calling conventions, etc., were Ada-ish and required
no understanding of anything outside Ada to use. (No pragma import stuff,
etc.) If it were legal for us mere mortal hackers to extend the package Ada,
it might have already been done. (This is how a lot of things crept into C,
after all.)

Its fair for someone to criticize Ada for not providing the sort of
reasonable and customary libraries one gets with lots of other languages.
(Where "reasonable and customary" stops is subject to debate, but I'd think
some basic OS services ought to fall within scope.) Saying "well you can get
there by binding to the C libraries" is a bit of a "me too!!!" syndrome that
keeps Ada a follower rather than a leader. Ada has to get out in front of
the issue and create its own interfaces or it will forever find itself
fighting the "well then why not just use C/C++ and be done with it?"
criticism. That's a hard one to win.

I like the idea of having Ada include lots of apendices to bind to things in
a standard way. A lot of what is in the C libraries Ada doesn't need because
it has other ways of getting there, but why not look over the C libraries
and provide an Ada standard way of getting the same services? (Not in a
slavish "me too!!!" way - do it "The Ada Way" (tm).)

A quick scan of K&R-II Appendix B, suggest the following as things Ada
provides no immediate alternative to that could be added in some way:

<stdlib.h> system
<stdlib.h> getenv
<stdlib.h> bsearch
<stdlib.h> qsort
<assert.h> (Yes, it exists, but should be part of the standard...)
<signal.h> (Yeah, you have some of it with exceptions - OS stuff is what I
have in mind here - can we have a standard exception (or other mechanism?)
for SIGINT, SIGTERM, SIGABRT?)
<time.h> clock
<time.h> asctime
<time.h> ctime
<time.h> gmtime
<time.h> localtime
<time.h> strftime

A little thought about other things commonly supplied by most OS's would
probably yield a bunch more ideas. Maybe we'd start seeing some posts in
Comp.Lang.C(++) to the effect of "I can do XYZ in Ada - how do I do the same
thing in C(++)?" and "Why doesn't C(++) provide me with a function/class to
do ABC which I can do in Ada so easily?" (Maybe we can start some shill
postings? :-)

Just an idea....

MDC
--
Marin David Condic
Senior Software Engineer
Pace Micro Technology Americas    www.pacemicro.com
Enabling the digital revolution
e-Mail:    marin.condic@pacemicro.com
Web:      http://www.mcondic.com/

Re: howto make system calls (newbie question)

[Marin David Condic]

Its nice to find myself being totally agreed with! :-) Maybe you can expand
on the idea. What sort of standard libraries would you like to see in Ada?

MDC
--
Marin David Condic
Senior Software Engineer
Pace Micro Technology Americas    www.pacemicro.com
Enabling the digital revolution
e-Mail:    marin.condic@pacemicro.com
Web:      http://www.mcondic.com/


"Beard, Frank" <beardf@spawar.navy.mil> wrote in message
news:mailman.989009884.30531.comp.lang.ada@ada.eu.org...
> I totally agree with Marin.
> Frank
>

Re: howto make system calls (newbie question)

[Samuel T. Harris]

Marin David Condic wrote:
> 
> Its nice to find myself being totally agreed with! :-) Maybe you can expand
> on the idea. What sort of standard libraries would you like to see in Ada?
> 

Of course, there are other "standards" which are not part
of the Ada standard. For instance, the POSIX Ada binding
has been invaluable to me in writing portable code which
requires facilities from the operating system.

Since this is standardized by IEEE, I feel any particular
need to include it as an Ada LRM annex.

-- 
Samuel T. Harris, Principal Engineer
Raytheon, Aerospace Engineering Services
"If you can make it, We can fake it!"

Re: howto make system calls (newbie question)

[Marin David Condic]

Well, not everything in the world needs to be in an annex to the ARM. The
Posix binding is a fair example. Of course, there you have the problem that
the interface may not be made available by a given compiler and it might not
be applicable in all environments anyway. That's a binding to a specific
class of OS that may not be common across all OS's & one would want to give
that some thought.

Is there a reason to maybe make an OS independent package available that
provided some of the things that might exist in Posix, but might be provided
by a different system in a different manner? Or is Posix widely spread
enough that Ada could just say "If you want access to system facilities of
this type - go see Posix"? Is the Posix binding going to just come with my
compiler and I can go along fat, dumb and happy using it at will with no
thought as to where to find it and no concern for transportability of my
code? The answers are not clear to me.

However, I still think there is room to add some standard libraries to Ada
that make it more useful to the programmer. Clearly the Ada.Strings...
branch is a good example - C may supply some fairly lame string handling
functions but Ada83 provided none at all. Now Ada95 has an area of clear
superiority in that you get some really good pre-fabricated string utilities
that are just there on any implementation and are at your disposal. Might
there not be similar libraries of things that Ada could provide in some
standardized manner? If a C programmer can say "I can do this thing in C -
how do I do it in Ada?" should it be fair game to ask how we might have that
provided in all Ada implementations?

I really wish the Ada Standard Library Working Group had managed to get
somewhere with defining even a starting point for this sort of thing. Even
if a library didn't come by virtue of the ARM, if it was available with a
standard interface and provided by most, if not all, implementations, it
would enhance the value of Ada to the programmer. Then we'd be off in other
newsgroups asking "I can do this in Ada - how do I do it in your language?"
:-)

MDC
--
Marin David Condic
Senior Software Engineer
Pace Micro Technology Americas    www.pacemicro.com
Enabling the digital revolution
e-Mail:    marin.condic@pacemicro.com
Web:      http://www.mcondic.com/


"Samuel T. Harris" <u61783@gsde.hou.us.ray.com> wrote in message
news:3AF6D040.FF71B782@gsde.hou.us.ray.com...

> Of course, there are other "standards" which are not part
> of the Ada standard. For instance, the POSIX Ada binding
> has been invaluable to me in writing portable code which
> requires facilities from the operating system.
>
> Since this is standardized by IEEE, I feel any particular
> need to include it as an Ada LRM annex.
>

Re: howto make system calls (newbie question)

[Samuel T. Harris]

Marin David Condic wrote:
> 
> Well, not everything in the world needs to be in an annex to the ARM. The
> Posix binding is a fair example. Of course, there you have the problem that
> the interface may not be made available by a given compiler and it might not
> be applicable in all environments anyway. That's a binding to a specific
> class of OS that may not be common across all OS's & one would want to give
> that some thought.

If I need POSIX, then that becomes a requirement for compiler choice.
POSIX compliance is also a common criteria for OS selection as well.

I also feel that POSIX is a good abstraction layer.

I believe most operating systems can be made to "fit" most
of the standard, at least when it comes to the most common
operations.

> 
> Is there a reason to maybe make an OS independent package available that
> provided some of the things that might exist in Posix, but might be provided
> by a different system in a different manner? Or is Posix widely spread
> enough that Ada could just say "If you want access to system facilities of
> this type - go see Posix"? Is the Posix binding going to just come with my
> compiler and I can go along fat, dumb and happy using it at will with no
> thought as to where to find it and no concern for transportability of my
> code? The answers are not clear to me.

POSIX come with Apex and VADS.

One can get Florist with GNAT, but I'm not sure of the
totality of support, especially when it comes to threads.

As to the most common operations, perhaps the approach
should be to define what is expected as the least common
denominator and standardize that in a POSIX way as what
every implementation is expected to support.

> 
> However, I still think there is room to add some standard libraries to Ada
> that make it more useful to the programmer. Clearly the Ada.Strings...
> branch is a good example - C may supply some fairly lame string handling
> functions but Ada83 provided none at all. Now Ada95 has an area of clear
> superiority in that you get some really good pre-fabricated string utilities
> that are just there on any implementation and are at your disposal. Might
> there not be similar libraries of things that Ada could provide in some
> standardized manner? If a C programmer can say "I can do this thing in C -
> how do I do it in Ada?" should it be fair game to ask how we might have that
> provided in all Ada implementations?

I'd like to see more work on numerics, especially matrix math.

> 
> I really wish the Ada Standard Library Working Group had managed to get
> somewhere with defining even a starting point for this sort of thing. Even
> if a library didn't come by virtue of the ARM, if it was available with a
> standard interface and provided by most, if not all, implementations, it
> would enhance the value of Ada to the programmer. Then we'd be off in other
> newsgroups asking "I can do this in Ada - how do I do it in your language?"

I do that whenever I can over at comp.software-eng and comp.object.

-- 
Samuel T. Harris, Principal Engineer
Raytheon, Aerospace Engineering Services
"If you can make it, We can fake it!"

Re: howto make system calls (newbie question)

[Marin David Condic]

On my web site there is a bunch of statistics code that was built with an
eye towards getting it into the Ada Standard Library Working Group output.
We already define stuff in an appendix for the more common log and trig
functions after discovering that Fortran weenies were complaining about them
being missing. Why not look to other mathematical things that can be
implemented in a fairly recognizable way? Matrix and vector math libraries
exist all over - producing something that would look familiar to those users
might be a useful extension. Obviously I and a couple of extremely bright
math guys thought that statistics would make a good addition. There must be
other math functions that would be useful as well. Lots of numeric
programmers might learn to love Ada as a result.

MDC
--
Marin David Condic
Senior Software Engineer
Pace Micro Technology Americas    www.pacemicro.com
Enabling the digital revolution
e-Mail:    marin.condic@pacemicro.com
Web:      http://www.mcondic.com/

"Samuel T. Harris" <u61783@gsde.hou.us.ray.com> wrote in message
news:3AF85750.D261FBC7@gsde.hou.us.ray.com...
>
> I'd like to see more work on numerics, especially matrix math.
>

Re: howto make system calls (newbie question)

[Samuel T. Harris]

"Samuel T. Harris" wrote:
> 
> Marin David Condic wrote:
> >
> > Its nice to find myself being totally agreed with! :-) Maybe you can expand
> > on the idea. What sort of standard libraries would you like to see in Ada?
> >
> 
> Of course, there are other "standards" which are not part
> of the Ada standard. For instance, the POSIX Ada binding
> has been invaluable to me in writing portable code which
> requires facilities from the operating system.
> 
> Since this is standardized by IEEE, I       feel any particular
                                       ^don't^
> need to include it as an Ada LRM annex.
> 

-- 
Samuel T. Harris, Principal Engineer
Raytheon, Aerospace Engineering Services
"If you can make it, We can fake it!"

Re: howto make system calls (newbie question)

[Tarjei T. Jensen]

Marin David Condic wrote in message <9d6c89$1nd$1@nh.pace.co.uk>...
>Its nice to find myself being totally agreed with! :-) Maybe you can expand
>on the idea. What sort of standard libraries would you like to see in Ada?

The obvious ones. E.g. console I/O, adafied sockets, buffered I/O, national
language support (NLS), turbo pascal style strings, NLS string comparison,
file utilities (expand wildcards, get file names from a directory, delete
file, set file protection, etc), dynamic library support (load library,
symbols under program control), standard datatypes for system calls,
standard datatypes for the C library.


Greetings,

Re: howto make system calls (newbie question)

[Larry Kilgallen]

In article <9d87id$oll15@news.kvaerner.com>, "Tarjei T. Jensen" <tarjei.jensen@kvaerner.com> writes:
> 
> Marin David Condic wrote in message <9d6c89$1nd$1@nh.pace.co.uk>...
>>Its nice to find myself being totally agreed with! :-) Maybe you can expand
>>on the idea. What sort of standard libraries would you like to see in Ada?
> 
> The obvious ones. E.g. console I/O, adafied sockets, buffered I/O, national
> language support (NLS), turbo pascal style strings, NLS string comparison,
> file utilities (expand wildcards, get file names from a directory, delete
> file, set file protection, etc), dynamic library support (load library,
> symbols under program control), standard datatypes for system calls,
> standard datatypes for the C library.

Console I/O - secure operating systems are likely to prevent a program
	run by an ordinary user from accessing the console, to prevent
	password grabbing attacks against the operator.

Turbo Poascal style strings - please describe.  Given Ada String,
Bounded String and Unbounded Strings, I am hard-pressed to thing
of capabilities that are missing.

Set File Protection - this certainly cannot be made uniform across
operating systems. VMS has SOGW, Unix (at least some) have one fewer.
A major way operating systems distinguish themselves is by differences
in their ACLs.

Standard datatypes for system calls - And what good will a VMS-style
descriptor do me on a Unix system ?

Re: howto make system calls (newbie question)

[Tarjei T. Jensen]

Larry Kilgallen
>Console I/O - secure operating systems are likely to prevent a program
> run by an ordinary user from accessing the console, to prevent
> password grabbing attacks against the operator.

Try again. console I/O means that there are no GUI.

>Turbo Poascal style strings - please describe.  Given Ada String,
>Bounded String and Unbounded Strings, I am hard-pressed to thing
>of capabilities that are missing.

This has been discussed before in this newsgroup.

>Set File Protection - this certainly cannot be made uniform across
>operating systems. VMS has SOGW, Unix (at least some) have one fewer.
>A major way operating systems distinguish themselves is by differences
>in their ACLs.

Of course it can. It is just a bit of work and som common sense. The posix
(or was it C) standard have some suggestions about how to do it.

>Standard datatypes for system calls - And what good will a VMS-style
>descriptor do me on a Unix system ?

Beats me. Why would you want that?


Greetings,

Re: howto make system calls (newbie question)

[Larry Kilgallen]

In article <9d8ute$8tt9@news.kvaerner.com>, "Tarjei T. Jensen" <tarjei.jensen@kvaerner.com> writes:
> 
> Larry Kilgallen
>>Console I/O - secure operating systems are likely to prevent a program
>> run by an ordinary user from accessing the console, to prevent
>> password grabbing attacks against the operator.
> 
> Try again. console I/O means that there are no GUI.

No, absent a specific explanation to the contrary, or specification
of some particular operating system, console I/O should mean I/O to
the console.  On the machines I use, ordinary unprivileged users are
_not_ allowed to write to the console (if they were, their output would
get intermixed).

>>Turbo Poascal style strings - please describe.  Given Ada String,
>>Bounded String and Unbounded Strings, I am hard-pressed to thing
>>of capabilities that are missing.
> 
> This has been discussed before in this newsgroup.

If you remember that, but don't remember the answer, don't bother
commenting on the question.

>>Set File Protection - this certainly cannot be made uniform across
>>operating systems. VMS has SOGW, Unix (at least some) have one fewer.
>>A major way operating systems distinguish themselves is by differences
>>in their ACLs.
> 
> Of course it can. It is just a bit of work and som common sense. The posix
> (or was it C) standard have some suggestions about how to do it.

And do they cover protected subsystem ACEs, which I need for VMS ?
I am sure that MVS, OS/400 and other operating systems have their
own unique ACE types, to say nothing of different processing rules.
As I said, it cannot be made universal across operating systems.

>>Standard datatypes for system calls - And what good will a VMS-style
>>descriptor do me on a Unix system ?
> 
> Beats me. Why would you want that?

The proposal was to implement standard datatypes for system calls.
A descriptor is the most common datatype for VMS system calls.  I
am sure that other operating systems have their own datatypes that
also do not have universal applicability.

If your intention is to provide a binding only to Posix-standard
interfaces, then say so, but I thought that already existed for Ada.

Re: howto make system calls (newbie question)

[Charles Hixson]

Well, if it were an optional annex, then it wouldn't be 
mandatorially present, but optional, as garbage collection is 
now.  (That may be a good argument against it, considering the 
status of garbage collection, however...)

Further:
Kilgallen@eisner.decus.org.nospam (Larry Kilgallen) wrote in
<Fpv1u1koJqhc@eisner.encompasserve.org>: 

>In article <9d8ute$8tt9@news.kvaerner.com>, "Tarjei T.
>Jensen" <tarjei.jensen@kvaerner.com> writes: 
>> 
>>...
>No, absent a specific explanation to the contrary, or
>specification of some particular operating system, console
>I/O should mean I/O to the console.  On the machines I use,
>ordinary unprivileged users are _not_ allowed to write to the
>console (if they were, their output would get intermixed).
>
Do you mean that all of your users are restricted to batch 
compile/execution cycles?  Then slave sysin to the card reader 
and sysout to the line printer (Unit5 and Unit6, I think they 
were called.)

...
>...
>And do they cover protected subsystem ACEs, which I need for
>VMS ? I am sure that MVS, OS/400 and other operating systems
>have their own unique ACE types, to say nothing of different
>processing rules. As I said, it cannot be made universal
>across operating systems. 

No.  But perhaps a higher level description could be.  It might 
not have the total functionality of low-level system calls, but 
one often doesn't need that. (I have only used VMS twice, so 
forgive me if I don't know what an ACE is.)

>
>...
>If your intention is to provide a binding only to
>Posix-standard interfaces, then say so, but I thought that
>already existed for Ada. 
>
Posix-standard interfaces are good.  That way at least one 
implementation has already been done, so it doesn't need to be 
redone.  It could be swallowed whole into an annex as a proposed 
implementation (licensing issues permitting).

-- 
Charles Hixson

Copy software legally, the GNU way!
Use GNU software, and legally make and share copies of software.
See http://www.gnu.org
    http://www.redhat.com
    http://www.linux-mandrake.com
    http://www.calderasystems.com/
    http://www.linuxapps.com/

Re: howto make system calls (newbie question)

[Larry Kilgallen]

In article <Xns909B95495DD5Fcharleshixsonearthli@207.217.77.25>, charleshixson@earthling.net (Charles Hixson) writes:

>>In article <9d8ute$8tt9@news.kvaerner.com>, "Tarjei T.
>>Jensen" <tarjei.jensen@kvaerner.com> writes: 
>>> 
>>>...
>>No, absent a specific explanation to the contrary, or
>>specification of some particular operating system, console
>>I/O should mean I/O to the console.  On the machines I use,
>>ordinary unprivileged users are _not_ allowed to write to the
>>console (if they were, their output would get intermixed).
>>
> Do you mean that all of your users are restricted to batch 
> compile/execution cycles?  Then slave sysin to the card reader 
> and sysout to the line printer (Unit5 and Unit6, I think they 
> were called.)

No, but users of interactive terminals (or emulators thereof)
do not have access to the _console_.  There is only one console
on the system.  There could be tens or hundreds of users (depending
on the size of the machine).  From the console one can issue hardware
commands to stop the machine, etc.

>>And do they cover protected subsystem ACEs, which I need for
>>VMS ? I am sure that MVS, OS/400 and other operating systems
>>have their own unique ACE types, to say nothing of different
>>processing rules. As I said, it cannot be made universal
>>across operating systems. 
> 
> No.  But perhaps a higher level description could be.  It might 
> not have the total functionality of low-level system calls, but 
> one often doesn't need that. (I have only used VMS twice, so 
> forgive me if I don't know what an ACE is.)

By ACE I mean "Access Control Entry", the fundamental building
block of an Access Control List.  I was under the impression
the same term was used for Unix and Windows.  It is the most
common method of complying with the C2 rules of the Orange Book.
While various operating systems have Access Control Lists
for file protection, the parsing rules differ from one OS to
another.  Some early Unix implementations came without ACLs,
but I think the need for C2 evaluation cured that for any
serious implementations.

Re: howto make system calls (newbie question)

[Charles Hixson]

Kilgallen@eisner.decus.org.nospam (Larry Kilgallen) wrote in
<FNjDhEG2MwD3@eisner.encompasserve.org>: 

>In article
><Xns909B95495DD5Fcharleshixsonearthli@207.217.77.25>,
>charleshixson@earthling.net (Charles Hixson) writes: 
>
>...
>> Do you mean that all of your users are restricted to batch 
>> compile/execution cycles?  Then slave sysin to the card
>> reader and sysout to the line printer (Unit5 and Unit6, I
>> think they were called.)
>
>No, but users of interactive terminals (or emulators thereof)
>do not have access to the _console_.  There is only one
>console on the system.  There could be tens or hundreds of
>users (depending on the size of the machine).  From the
>console one can issue hardware commands to stop the machine,
>etc. 

OK.  In that case let it be re-stated as a request for access to 
terminal screen I/O.  I believe that this is what was meant.  (I 
know few people who have a separate system console anymore, so 
that meaning has rather dropped from my vocabulary [of course, 
I'm not the original poster, so I'm guessing what was meant].)
>...
>> 
>> No.  But perhaps a higher level description could be.  It
>> might not have the total functionality of low-level system
>> calls, but one often doesn't need that. (I have only used
>> VMS twice, so forgive me if I don't know what an ACE is.)
>
>By ACE I mean "Access Control Entry", the fundamental
>building block of an Access Control List.  I was under the
>impression the same term was used for Unix and Windows.  It
>is the most common method of complying with the C2 rules of
>the Orange Book. While various operating systems have Access
>Control Lists for file protection, the parsing rules differ
>from one OS to another.  Some early Unix implementations came
>without ACLs, but I think the need for C2 evaluation cured
>that for any serious implementations.
>
OK.  Then perhaps the difference is that I never work with 
highly secured systems.  Perhaps that *is* specialized enough 
(and different enough from system to system) that a high level 
specification would be inappropriate.  OTOH, if a similar 
construct is used on VMS, *nix, and Windows, then perhaps a high 
level description would be appropriate.

-- 
Charles Hixson

Copy software legally, the GNU way!
Use GNU software, and legally make and share copies of software.
See http://www.gnu.org
    http://www.redhat.com
    http://www.linux-mandrake.com
    http://www.calderasystems.com/
    http://www.linuxapps.com/

Re: howto make system calls (newbie question)

[Larry Kilgallen]

In article <Xns909C5BB41B746charleshixsonearthli@207.217.77.22>, charleshixson@earthling.net (Charles Hixson) writes:
> Kilgallen@eisner.decus.org.nospam (Larry Kilgallen) wrote in
> <FNjDhEG2MwD3@eisner.encompasserve.org>: 
> 
>>In article
>><Xns909B95495DD5Fcharleshixsonearthli@207.217.77.25>,
>>charleshixson@earthling.net (Charles Hixson) writes: 

>>> No.  But perhaps a higher level description could be.  It
>>> might not have the total functionality of low-level system
>>> calls, but one often doesn't need that. (I have only used
>>> VMS twice, so forgive me if I don't know what an ACE is.)
>>
>>By ACE I mean "Access Control Entry", the fundamental
>>building block of an Access Control List.  I was under the
>>impression the same term was used for Unix and Windows.  It
>>is the most common method of complying with the C2 rules of
>>the Orange Book. While various operating systems have Access
>>Control Lists for file protection, the parsing rules differ
>>from one OS to another.  Some early Unix implementations came
>>without ACLs, but I think the need for C2 evaluation cured
>>that for any serious implementations.
>>
> OK.  Then perhaps the difference is that I never work with 
> highly secured systems.  Perhaps that *is* specialized enough 
> (and different enough from system to system) that a high level 
> specification would be inappropriate.  OTOH, if a similar 
> construct is used on VMS, *nix, and Windows, then perhaps a high 
> level description would be appropriate.

If there were a uniform syntax, I worry that some might write
Ada programs that were apparently portable (they compiled and
executed on multiple platforms) but in fact had vastly different
effective file protections depending on the operating system.
I think one of the underlying principles of Ada is to avoid
leading the programmer into mistaken practices.

Re: howto make system calls (newbie question)

[Tarjei T. Jensen]

Larry Kilgallen wrote
>The proposal was to implement standard datatypes for system calls.
>A descriptor is the most common datatype for VMS system calls.  I
>am sure that other operating systems have their own datatypes that
>also do not have universal applicability.

Why would someone include VMS specifics for Unix or windows nt for that
matter? It makes no sense to have items specific for Unix on a VMS system
(unless one is using the posix subsystem).


Greetings,

Re: howto make system calls (newbie question)

[Larry Kilgallen]

In article <9dauu7$8m913@news.kvaerner.com>, "Tarjei T. Jensen" <tarjei.jensen@kvaerner.com> writes:
> 
> Larry Kilgallen wrote
>>The proposal was to implement standard datatypes for system calls.
>>A descriptor is the most common datatype for VMS system calls.  I
>>am sure that other operating systems have their own datatypes that
>>also do not have universal applicability.
> 
> Why would someone include VMS specifics for Unix or windows nt for that
> matter? It makes no sense to have items specific for Unix on a VMS system
> (unless one is using the posix subsystem).

I certainly agree, but I was responding to a thread with entries
espousing codification of all "standard datatypes" that might be
used in system calls.  I consider that unworkable, because what
constitutes a "standard datatype" varies so much between various
operating systems.   Basically, I do not feel there is progress
to be made in that direction.

Re: howto make system calls (newbie question)

[Charles Hixson]

Kilgallen@eisner.decus.org.nospam (Larry Kilgallen) wrote in
<2ecbjJt$+0y8@eisner.encompasserve.org>: 

>In article <9dauu7$8m913@news.kvaerner.com>, "Tarjei T.
>Jensen" <tarjei.jensen@kvaerner.com> writes: 
>...
>I certainly agree, but I was responding to a thread with
>entries espousing codification of all "standard datatypes"
>that might be used in system calls.  I consider that
>unworkable, because what constitutes a "standard datatype"
>varies so much between various operating systems.  
>Basically, I do not feel there is progress to be made in that
>direction. 
>
Perhaps the problem is the correct level of abstraction.  
Clearly machine/OS specific variations don't belong in the 
standard, but I also feel that SOME way of accomplishing the 
needed tasks does belong in the standard.  It may be that on 
some machines this would be inappropriately inefficient, but at 
least it would be possible.  This is somewhat similar to using a 
numeric 1 to indicate true.  If your machine has a 60 bit word 
length, then this may be quite inefficient.  But it will make 
the transmission of logical information possible.  (I believe 
that this was, indeed, the way that it was done on the CDC7600, 
at least in the Fortran that I used.)

As with the representation of Booleans, so with other things, it 
is frequently true that a less efficient method will be "good 
enough", and that portability will pay back for numerous 
inefficiencies.  And a good compiler could come with specialized 
libraries that would handle the process in a way more 
appropriate to the particular machine/OS.  (On the 7600, it was 
a speed trade off.  The individual bits couldn't be addressed, 
so for speed the >0 comparison was used [I believe that other 
compilers just checked the sign bit]).

-- 
Charles Hixson

Copy software legally, the GNU way!
Use GNU software, and legally make and share copies of software.
See http://www.gnu.org
    http://www.redhat.com
    http://www.linux-mandrake.com
    http://www.calderasystems.com/
    http://www.linuxapps.com/

Re: howto make system calls (newbie question)

[Tarjei T. Jensen]

Larry Kilgallen wrote
>I certainly agree, but I was responding to a thread with entries
>espousing codification of all "standard datatypes" that might be
>used in system calls.  I consider that unworkable, because what
>constitutes a "standard datatype" varies so much between various
>operating systems.   Basically, I do not feel there is progress
>to be made in that direction.


As long as that bit is system specific, it will work as intended. VMSers do
not have this problem as most if not all of the stuff needed to interact
with the system is already present. For Unix, windows nt, etc this is not
the case.

If a program uses sockets which are common to quite a lot of operating
systems, it will work on all if it uses the definitions from a platform
specific source. Granted that sockets and related routines are not
neccessarily a system resource in a given operating system, but they are
pervasive enough to be included if they are supported on the platform.


Greetings,

Re: howto make system calls (newbie question)

[Marin David Condic]

I like the Ada-fied sockets part in particular. It would be nice to get
bindings to common things that used Ada-ish parameters. (Sockets made to
line up nicely with Ada.Streams? Hmmmmm???? A nice Object Oriented approach
to sockets rather than the customary C-ish approach?)

MDC
--
Marin David Condic
Senior Software Engineer
Pace Micro Technology Americas    www.pacemicro.com
Enabling the digital revolution
e-Mail:    marin.condic@pacemicro.com
Web:      http://www.mcondic.com/


"Tarjei T. Jensen" <tarjei.jensen@kvaerner.com> wrote in message
news:9d87id$oll15@news.kvaerner.com...
>
> Marin David Condic wrote in message <9d6c89$1nd$1@nh.pace.co.uk>...
> >Its nice to find myself being totally agreed with! :-) Maybe you can
expand
> >on the idea. What sort of standard libraries would you like to see in
Ada?
>
> The obvious ones. E.g. console I/O, adafied sockets, buffered I/O,
national
> language support (NLS), turbo pascal style strings, NLS string comparison,
> file utilities (expand wildcards, get file names from a directory, delete
> file, set file protection, etc), dynamic library support (load library,
> symbols under program control), standard datatypes for system calls,
> standard datatypes for the C library.
>
>
> Greetings,
>
>
>

Re: howto make system calls (newbie question)

[Randy Brukardt]

Marin David Condic wrote in message <9d8t6a$t8r$1@nh.pace.co.uk>...
>I like the Ada-fied sockets part in particular. It would be nice to get
>bindings to common things that used Ada-ish parameters. (Sockets made
to
>line up nicely with Ada.Streams? Hmmmmm???? A nice Object Oriented
approach
>to sockets rather than the customary C-ish approach?)

Already done. Take a look at Claw.Sockets (a subset is available in the
Claw Introductory Version, which is a free download from
www.rrsoftware.com).

            Randy.

Re: howto make system calls (newbie question)

[Larry Kilgallen]

In article <pE1L6.1995$A07.421101@homer.alpha.net>, "Randy Brukardt" <randy@rrsoftware.com> writes:
> Marin David Condic wrote in message <9d8t6a$t8r$1@nh.pace.co.uk>...
>>I like the Ada-fied sockets part in particular. It would be nice to get
>>bindings to common things that used Ada-ish parameters. (Sockets made
> to
>>line up nicely with Ada.Streams? Hmmmmm???? A nice Object Oriented
> approach
>>to sockets rather than the customary C-ish approach?)
> 
> Already done. Take a look at Claw.Sockets (a subset is available in the
> Claw Introductory Version, which is a free download from
> www.rrsoftware.com).

I _think_ (without studying it) I like the Claw licensing approach
(more traditional) better than xGPL, but I was under the impression
that Claw was only for Microsoft operating systems.

I thought this discussion was looking for general Ada interfaces,
which would make me say the unseen (by me) Claw approach would at
best be a "worked example" rather than an "already done".

Ada Libraries (was Re: howto make system calls (newbie question))

[Marin David Condic]

It may already be done in Claw, which is great for Windows based stuff, but
doesn't help much for non windows systems. Originally this came up as a
discussion about some sort of standardized library of stuff that would
generally be available to all Ada compilers. Sort of like being able to
count on the class libraries in Java. (Data structures, math stuff,
communications, etc. Anything that could reasonably be expected to work on
most systems to which Ada is targeted.) The question would be, how easily
would the Claw.Sockets stuff port to - say - a Linux environment? And what
is the liklihood that Claw.Sockets could be bundled up with a bunch of other
things and put under some sort of GPL-ish license so that compiler vendors
would be free to distribute it with their product?

BTW: I liked the Ada Developers Cooperative License proposed by Bob Leif as
a means by which an individual or small company could make Ada code
available at no charge, but could lay claim to revenue if the code is
incorporated in a product for resale. I think it would need to come up with
some sort of payment scheme beyond a straight percentage of gross sales, and
a lot of details would need to be hammered out, but in general, I think the
concept is a good one.

MDC
--
Marin David Condic
Senior Software Engineer
Pace Micro Technology Americas    www.pacemicro.com
Enabling the digital revolution
e-Mail:    marin.condic@pacemicro.com
Web:      http://www.mcondic.com/


"Randy Brukardt" <randy@rrsoftware.com> wrote in message
news:pE1L6.1995$A07.421101@homer.alpha.net...
> Already done. Take a look at Claw.Sockets (a subset is available in the
> Claw Introductory Version, which is a free download from
> www.rrsoftware.com).
>

Re: Ada Libraries (was Re: howto make system calls (newbie question))

[Randy Brukardt]

Marin David Condic wrote in message <9domjk$h4t$1@nh.pace.co.uk>...
>It may already be done in Claw, which is great for Windows based stuff,
but
>doesn't help much for non windows systems. Originally this came up as a
>discussion about some sort of standardized library of stuff that would
>generally be available to all Ada compilers. Sort of like being able to
>count on the class libraries in Java. (Data structures, math stuff,
>communications, etc. Anything that could reasonably be expected to work
on
>most systems to which Ada is targeted.) The question would be, how
easily
>would the Claw.Sockets stuff port to - say - a Linux environment? And
what
>is the liklihood that Claw.Sockets could be bundled up with a bunch of
other
>things and put under some sort of GPL-ish license so that compiler
vendors
>would be free to distribute it with their product?

Sockets stuff isn't very Windows specific; the implementation would of
course need to be redone for other systems.

As far as the license goes, the question is more whether the license
would allow vendors to distribute it with their compilers. Indeed, I
would expect a GPL license to PREVENT most of the vendors from
distributing it. (GPL poisoning, anyone?)

It was always our intent that Claw be distributed with compilers, indeed
it is distributed with Janus/Ada, and the introductory version is
distributed with Rational Apex and it also is on the ASE CD-ROM (as well
as our website).

In any case, I was responding more to the notion that "someone needs to
design" a O-O sockets library -- that's been done. Not that Claw.Sockets
as it stands would actually be the answer to everyone's needs.

Hopefully, this will be one of the areas that will be addressed in Ada
0y. As directory operations already is.

                Randy Brukardt.

Re: Ada Libraries

[Marin David Condic]

O.K. Just to be fair - I did say "GPL-ish". :-) I'd hate to start that whole
GPL/LGPL/GMGPL thing all over again. I was only observing that if someone
were to develop something that hoped to make its way into Ada as a library,
there needs to be some means that makes it very easy for a compiler vendor
just to bundle it up and ship it. (That and it has to be good quality,
general purpose & portable. We could go on for days about all the possible
requirements. For now I'll stick to the licensing issue.)

MDC
--
Marin David Condic
Senior Software Engineer
Pace Micro Technology Americas    www.pacemicro.com
Enabling the digital revolution
e-Mail:    marin.condic@pacemicro.com
Web:      http://www.mcondic.com/


"Randy Brukardt" <randy@rrsoftware.com> wrote in message
news:MaWL6.2162$A07.433578@homer.alpha.net...
> Marin David Condic wrote in message <9domjk$h4t$1@nh.pace.co.uk>...
> >things and put under some sort of GPL-ish license so that compiler
> vendors
> >would be free to distribute it with their product?
>
>
> As far as the license goes, the question is more whether the license
> would allow vendors to distribute it with their compilers. Indeed, I
> would expect a GPL license to PREVENT most of the vendors from
> distributing it. (GPL poisoning, anyone?)
>

Re: Ada Libraries (was Re: howto make system calls (newbie question))

[tmoran]

> The question would be, how easily would the Claw.Sockets stuff port
> to - say - a Linux environment?
  My understanding is that winsock is pretty similar to Unix sockets,
so the translation there shouldn't be hard.  Claw.Sockets of course
uses the Windows mechanisms to be notified of asynchronous events, so
that would have to change to Linux mechanisms.  The Ada-ification
stuff, parameters, exceptions, controlled types, protected objects,
etc, could presumably remain unchanged.  I suppose there could be some
whole areas of functionality that wouldn't port well.  The only thing
that springs to mind is the absence of asynchronous sockets in WinCE,
but my understanding is they were used mostly in older, single-tasking
Windows 3.1 environments.

Ada Libraries

[Randy Neff]

Read  "Source Code" by Mark Weisner of Xerox PARC, in IEEE Computer, Nov 87.

This paper is an excellent discussion of why programmers should have reading
access to all source code.   My twenty odd years of programming experience
agrees with his arguements.   I.E. writing an Ada interface to Un*x curses
required reading the source code;  a complete Ada interface to Sun CGI 
was aborted because no source was available.

I have wasted weeks of time with buggy Ada compilers when the error 
messages (like internal assertion error 59) were meaningless.   Having the
source code and debugger would at least allow me to find out what caused
the error and how to write a work around.   No way will I fix the compiler;
it is just that the error messages and documentation are useless.

However, if a vendor wants to cripple the programmers that buy its library by
not providing source code, then here is one note:   the Verdix compiler
library has the source file name, which is used by the debugger.  
Delete or rename the source file and the debugger cannot find it.

Motherhood:   One of the big unanswered questions in CS is how to 
define the word 'guarantee' as in:  "I, the vendor, guarantee that the
specifications are complete, correct, and consistent; and that the 
hidden code I provide is complete, correct, and consistent with the 
specifications."   Unfortunately,  today's state of the art is that the
only trustworthy specification is the machine code; if you trust your
assembler, then the assembly code; if you trust your Ada compiler (I don't),
then the Ada source code.   

Randall Neff @sierra.stanford.edu

Ada Libraries

[amiram]

Is there any standard and/or recommended spec for what needs to be
included in an Ada Library system, and how it should be represented?
Any agreement among compiler writers about it?
Amiram Yehudai.