Re: The TTN dilemma (was Re: guile-www)

[Chris Hall]

Thamer Al-Harbash <address@hidden> writes:

> On Thu, 3 Jun 2004, Chris Hall wrote:
>
>> Oh, really?  Could you tell me a bit more about this?
>
> Sure. I'm hoping it will solicit a larger response from the guile
> community on what can be done about this.

Thanks for taking the time, I've been very curious about this.

How long has this been going on, roughly speaking?  Is there a thread
somewhere that I could read to catch up on the discussion?

>
> Some time ago, I found that guile-www from the savannah
> distribution had a small bug with optional arguments. You can
> probably find my patch in the guile-bugs archive. I received two
> responses. One from Marius thanking me, and one from TTN saying
> something else. Marius also asked TTN to stop representing his
> fork as the official one.

By 'his fork', do you mean 'his fork' of guile-pg, 'his fork' of
guile-www, etc.?

The little bit I can glean as to the 'why' of the fork from TTN's 'glug'
web page is that 'the fork' was *because* of a backward compatability
issue?

Does anyone else find it ironic (as well as a PITA) that this seems to
have led to many more, but undocumented, incompatabilities?

>
> Basically if you get modules from TTN they may not work with the
> GNU guile distribution, an example is guile-sdl. The response
> from TTN is to get the 'other' guile to work with his code and
> the response from the guile developers hasn't been vocal enough
> for me to understand.

Do the guile-pg-0.19, and perhaps the recent guile-www as well, require
a 'TTN-built' tool chain of some sort?

At least, the guile-pg install - it seemed to build OK - was *totally*
broken - it created directories off the '/' dir instead of where they
were _supposed_ to go, and said they would go, and then couldn't find a
new 'postgres-sup' lib, even though it had just installed it!

From what I can see in the guile-pg-0.19 distribution materials, there
is absolutely no mention of possible imcompatabilites, nor on the web
site where I got the tarball - I just checked again.  There is a _lot_
of install-related stuff, but no mention of incompatability that I could
see.

For me thi install was a _very_ lengthy and _very_ exasperating
experience - I nearly gave up on guile (again) because what good is an
extension language with broken extensions?  How useful are extensions
that it takes longer to install than it would take to write a
'quick-n-dirty' that could then evolve over time?

I am also beginning to get the feeling that ttn is a 'community of one'
- is this correct?

>
> There was an attempt to merge the code bases by someone on the
> list. No idea where it went.
>
> My personal opinion on this is that TTN ought to use his
> productivity to further guile development and not splinter it
> like he's doing now.

I *do* think that it is wonderful that ttn has been so active and is
contributing back - my contributions to the community are pitiful in
comparison.

But what issue is so important that splitting off a fork of a tool that
is by it's very nature meant to be unifying is the best route to take?

Quite frankly, I _do_ find it a bit deceptive to not say right up front
that 'this is not the official GNU/FSF stuff' if in fact it isn't, and I
have to wonder as to why it doesn't.  And at the bottom of the Guile-PG
page it says:

        There is one known bug we can't do anything about (we tried but
        failed): The sourceforge folks don't want to update their site
        to point to the new Guile-PG homepage:

            http://www.glug.org/people/ttn/software/guile-pg/

        But if there are enough links to the new page, perhaps the
        relevancy of sourceforge's obstinance will decline over time.

Could the reason be that the people that actually *registered* the
project and thus have the *rights* to the page at sourceforge didn't see
any reason to change the link?  And again, I am left wondering where the
rest of the 'we' is, unless ttn is using the 'royal we'? ;-)

On the 'index' page for ttn's free software it mentions that several
projects were 'inherited from such-and-so'.

Hmmm.  Does that mean the project originators in some way assigned the
responsibility to ttn?  Or (heaven forbid!)  they passed away and willed
the responsibility to ttn?

My understanding of how these things usually go is that there is some
sort of hand-off done publicly in full 'view' of the community - has
this happened with the 'inherited' projects, I wonder?

>
> Unfortunately the Free Software licenses can only promote
> cooperation. They can't force it.

'Forced co-operation'?  Heh.  Surely you jest?  ;-D

Well, I am getting ready to move, but I have spent so much time on the
whole experience that maybe when I get settled in at the new place I'll
take whatever I like of ttn's guile-pg stuff and see about sending a
diff to the original maintainers.  As long as the API's aren't too
different, it may be the best thing.  Or maybe just update the original
to use the newer postgres headers/libs - that seems to be the issue that
caused me the problems in the first place.

Duplication of effort.  Blehhh.  Too much like work.

Aloha,
+Chris

-- 
There are 10 kinds of people in the world.  Those who understand
binary, and those who don't.
-- Unknown

pgpxSotcvpG47.pgp
Description: PGP signature