Re: ANN: Basil -- Internet Message (email) and MIME library for Ada v 1.0

[google1@hafdconsulting.com]

On Jul 22, 12:19 pm, Ludovic Brenta <ludo...@ludovic-brenta.org>
wrote:
> If you're planning to use AWS's SMTP client code, it looks like you're
> going to need this glue for your mailing list server anyway.

SMTP won't be needed for the mailing list. People want mailing list
servers
that are integrated with their existing MTA systems, they don't want
stand-alone systems. Mailman, one should note, provides interfaces
with
all the popular Unix MTAs and it's quite a bit of their code base,
they
did not, last I looked, provide an integrated MTA. There *might* be
some advantage to providing an integrated MTA particularly if AWS
makes
that very easy to do, but it would not be a very popular feature of
the
system at all.

I am considering using AWS for the web interface portion, but then
again
you have the same integration problem. People want their web
applications
to run behind the web server *they* choose (or at least Apache) they
don't
want a stand-alone system. There are some people around working on web
interface systems for Ada. When I get to that phase I'll look at them,
and
also look into the possibility of making an AWS-based system integrate
with
Apache. But integration with existing web servers is a killer feature
for
any interface I choose.

> The reason I wasn asking is not because I was interested in using your
> library; only because I think it is nice in general to consolidate
> small but related libraries into larger ones,

I take exception to your belief that Basil is a "small" library:

[murat:~/projects/basil-1.0]$ sloccount .
[...]
Totals grouped by language (dominant language first):
ada:          10823 (98.29%)
python:         146 (1.33%)
sh:              42 (0.38%)

Total Physical Source Lines of Code (SLOC)                = 11,011
Development Effort Estimate, Person-Years (Person-Months) = 2.48
(29.79)
 (Basic COCOMO model, Person-Months = 2.4 * (KSLOC**1.05))
Schedule Estimate, Years (Months)                         = 0.76
(9.08)
 (Basic COCOMO model, Months = 2.5 * (person-months**0.38))
Estimated Average Number of Developers (Effort/Schedule)  = 3.28
Total Estimated Cost to Develop                           = $ 335,397
 (average salary = $56,286/year, overhead = 2.40).
SLOCCount, Copyright (C) 2001-2004 David A. Wheeler

> or at least host them in
> common places, so that users can find them more easily.

Well, I have a 'selfish asshole' motive for working on these projects
and provide them as free software, that is to promote my business.
And that is why they'll be hosted on my professional website.
But I think inside a week if you put "MIME" and "Ada" into google,
that page will be the top hit, with this message and other forms of
publicity. Mr. Hermann has already informed me that he has added it
to his list of Ada resources, for instance. And of course, the Debian
developers around here could certainly help make it easier for people
to find and use it. :)

> Another
> benefit is to avoid inconsistencies and incompatilibities between
> libraries, so that a single program can use several libraries
> simultaneously.  And, of course, to reduce duplication of work.
> Speaking of which, is there a good reason why Basil needs its own
> "header objects" different from AWS's "header objects"?

Compared to *everthing else* in RFC 2822 and MIME, unstructured
headers are a very simple concept. They're just a key string, a colon,
and a value string, terminated by a non-folded CRLF sequence. RFC
2822 specifies the CRLF but to make Basil more robust, I made it also
handle lone CR or LFs.

So that was no great part of the library but it is a core part. To
use AWS I would have had to integrate every layer with that system,
making it a hard-dependency for building or using Basil, just so it
could do the light lifting beneath all of Basil's heavy lifting.

Also, my header lists *are* integrated into the Doubly_Linked_List
containers. My Headers.Lists.List object is a subtype of the DLL,
so any operation you could perform on a DLL you can perform on my
lists (except sorting isn't supported because of that interior
generic). I also provide a custom equality function that considers
things equal if their case insensitive keys are equal (per RFC 2822),
this allows you do do something like use the Find operation to iterate
over all headers with the same key.

I tried to make this library integrate very well with the Ada
language,
to leverage the power of the features that it has and provide an
interface that mimics the behavior of relevant core interfaces that
are already well-understood. I didn't see much use in integrating
with
vaguely related external third-party libraries, making the system
more
difficult to install and use.

But anyway, that's why I made the technical decisions I made. Of
course
writing software is the process of choosing between many different
choices which all have their merits and downsides. Maybe you're right
and I should have done this completely differently, there's not really
any way to know for sure. I would like to thank you for your
incredibly
fast response to the bug I filed against gnatpp.

Thanks,

--
Jordan Bettis