Homo or hetero

Homo or hetero

[Daniel Wengelin]

Homo or hetero?

I have seen the GNAT for RT-Linux press release. Whenever it is
released, it will be a very interesting product for different projects
I�m affiliated with. However, it leaves me wondering about one thing.

According to my (crude) definition of a homogenous distributed system,
it basically allows data to be correctly interpreted from the native
representation on any node in the distributed system. If data has to be
transformed to a neutral format to be correctly interpreted after a
transfer between nodes, the system is heterogeneous. A homogenous system
allows programs to pass data as it is represented natively between
programs without transformation, which has some obvious benefits and
drawbacks.

So, is a distributed system based on Linux and RT-Linux nodes running
GNAT Ada programs a homogenous system?

Obviously, there are many qualifications to be made. Please, if there is
any light to be shed on the issue!

Rgds, Daniel

PS. Please mail me directly if you can, I don�t come around to reading
the list more than once a month :-(

PPS Why does my spell checker say �homogenous� is OK; as is
�homogeneous�; �heterogenous� is not OK, it only allows �heterogeneous�?
;-)

Re: Homo or hetero

[Oliver Kellogg]

In article <38B2349E.83438A36@celsiustech.se>,
  Daniel Wengelin <dawe@celsiustech.se> wrote:
>
> So, is a distributed system based on Linux and RT-Linux nodes running
> GNAT Ada programs a homogenous system?

If the implication is [... Linux and RT-Linux] "running on processors
of same architecture" (for example, x86) then the answer should be
yes.
Anything else (for example, different processors albeit of same
endianness) would not necessarily constitute a homogeneous system.
Anyone from ACT listening?

-- Oliver M. Kellogg
-- okellogg at freenet dot de


Sent via Deja.com http://www.deja.com/
Before you buy.

Re: Homo or hetero

[Robert Dewar]

In article <89eda6$7vn$1@nnrp1.deja.com>,
  Oliver Kellogg <okellogg@my-deja.com> wrote:
> Anyone from ACT listening?


Sure, but there does not seem to be anything to add here. This
seems to be simply a discussion of relatively unimportant
terminology issues, or did I miss something?

Robert Dewar
Ada Core Technologies


Sent via Deja.com http://www.deja.com/
Before you buy.

Re: Homo or hetero

[Oliver Kellogg]

In article <89ee5l$8in$1@nnrp1.deja.com>,
  Robert Dewar <robert_dewar@my-deja.com> wrote:
>
> Sure, but there does not seem to be anything to add here. This
> seems to be simply a discussion of relatively unimportant
> terminology issues, or did I miss something?

Okay, I guess those spoiled by the joys of CORBA could call
the addressing of heterogeneousness by the DSA "unimportant"
 :-)

-- Oliver


Sent via Deja.com http://www.deja.com/
Before you buy.

Re: homo or hetero

[Oliver Kellogg]

In article <89i5vb$tqp$1@nnrp1.deja.com>,
  Robert Dewar <robert_dewar@my-deja.com> wrote:

> [...], but what possible effect can it have on the use of the
> DSA whether you decide to call this situation homogenous or
> heterogenous?

The confusion stems from the Annex E implementation
permission that

"An implementation may require that the set of processing
 nodes of a distributed system be homogeneous."

The operating system (e.g. Linux vs. Linux/RT) *should*
have no effect on the definition of "homogeneous" ?

-- Oliver
-- okellogg at freenet dot de


Sent via Deja.com http://www.deja.com/
Before you buy.

Re: homo or hetero

[Robert Dewar]

In article <89j0hp$fl3$1@nnrp1.deja.com>,
  Oliver Kellogg <okellogg@my-deja.com> wrote:
> In article <89i5vb$tqp$1@nnrp1.deja.com>,
>   Robert Dewar <robert_dewar@my-deja.com> wrote:
>
> > [...], but what possible effect can it have on the use of
the
> > DSA whether you decide to call this situation homogenous or
> > heterogenous?
>
> The confusion stems from the Annex E implementation
> permission that
>
> "An implementation may require that the set of processing
>  nodes of a distributed system be homogeneous."
>
> The operating system (e.g. Linux vs. Linux/RT) *should*
> have no effect on the definition of "homogeneous" ?


This term is not defined in the RM, hence it is implementation
defined, so the issue is not whether something is homogenous
or not, but whether the particular configuration is supported
by the implementation you are using of Annex E. Pretty clearly
an implementation might or might not support two different
operating systems (i.e. might consider them homogenous).

But as I said earlier, the discussion is entirely academic,
since the only available implementation of Annex E,
provided as the GLADE product that works with GNAT, fully
handles the heterogenous case anyway and does not take
advantage of this very ill-defined implementation permission.
(the implementation permission is a bit silly anyway, seeing
as there is permission to omit the entire annex :-)

Robert Dewar
Ada Core Technologies


Sent via Deja.com http://www.deja.com/
Before you buy.

Re: Homo or hetero

[Robert Dewar]

In article <89fol9$6nr$1@nnrp1.deja.com>,
  Oliver Kellogg <okellogg@my-deja.com> wrote:

> Okay, I guess those spoiled by the joys of CORBA could call
> the addressing of heterogeneousness by the DSA "unimportant"
>  :-)

Sorry, I am still completely confused. I know there is a smiley
there, but what possible effect can it have on the use of the
DSA whether you decide to call this situation homogenous or
heterogenous? Sure you have to check that your vendor's
implementation of Annex E will handle this combination, but
since so far the only implementation of Annex E is GNAT, and
GNAT handles heterogenous cases as easily as homogenous cases,
even if the architectures involved are very different (unlike
this case), that observation is academic!

Robert Dewar
Ada Core Technologies


Sent via Deja.com http://www.deja.com/
Before you buy.