Re: If i had a hammer...

Re: If i had a hammer...

[Markus Mottl]

Hello,

> The Markus Mottl message have revealed me that it could be good to have
> a page presenting the software developping under Caml and also projects
> under construction.. with
>       - a short description,
>       - name of the author(s) with mail, link to home page and other
>       stuff,
>       - last version / a progress indicator if it is a project
>       - .. snapshots.. every thing that could have an interest..
> what do you think about ?
> I can find a place for it (but perhaps it is better on the Inria
> web-site/ Caml page (i do not if it is possible..)..) and beginning the
> job..

I would really like to see a remote CVS-repository for OCAML somewhere. I
am sure this would very strongly help people contribute collectively
to projects written in OCAML. I have already asked at our university,
but people there are a bit reluctant with software that they don't know
so well and which allows remote access.

Looking at other projects which make excessive use of this tool shows,
how productive it can make people. It also attracts a lot of them -
I think this would be really important for the whole OCAML-project.

Is there any machine at INRIA that could be used for CVS? With read-only
access to the repository for everyone and write permissions for people
who have already contributed useful code to a project?

Best regards,
Markus

--
Markus Mottl, mottl@miss.wu-wien.ac.at, http://miss.wu-wien.ac.at/~mottl

Re: If i had a hammer...

[Gerd Stolpmann]

On Thu, 28 Jan 1999, Markus Mottl wrote:

>I would really like to see a remote CVS-repository for OCAML somewhere. I
>am sure this would very strongly help people contribute collectively
>to projects written in OCAML. I have already asked at our university,
>but people there are a bit reluctant with software that they don't know
>so well and which allows remote access.
>
>Looking at other projects which make excessive use of this tool shows,
>how productive it can make people. It also attracts a lot of them -
>I think this would be really important for the whole OCAML-project.
>

A public repository would be great. You can simply see what's currently
happening at a glance; it shows that there is really a community that
favours Ocaml.

I think there are several steps:

- Find a server where the projects can be put 

- Write a Web interface to simplify download

- Define a standard for documentation (very important for free software
  development) -- we can simply use an existing one such as 'info'
  or 'perldoc', or simply define a directory structure for html-pages

- Define a standard how to compile and install projects such that they
  can be used together

For the latter, I think a meta-makefile mechanism would be appropriate.
As Ocaml links always statically, there should be a tool

- that has as input a descriptive request what modules should be linked in

- that outputs a perfect Makefile that knows the linking order and automatically
  calls ocamlc/ocamlopt/ocamlmktop -- whatever meets the given requests best

For example, you could *describe* your request as follows:

	require Unix
	require Joe's_module
	require Mary's_module
	goal my_binary
	link my_binary = <requested libraries> m1.cmo m2.cmo

The meta-makefile processor reads this, searches in its database of already
installed modules what files are really meant by the given requests, orders
them such that they can be linked, and writes a Makefile that compiles 
m1.ml to m1.cmo, m2.ml to m2.cmo and links all the cmo and cma files.

Note that a similar description could be given how to compile Joe's and Mary's
modules. These can of course require third modules that must also be linked in.
My suggestion addresses both installation of downloaded modules and how to
use them.

The really important point is that using modules found in a public repository
must be very simple. It would discourage potential users if they had to
read manuals very carefully to get the modules working. It should not be
necessary to study and configure makefiles to do a simple thing like a
software installation (people often think that installation is a simple
process and a more or less an unproductive task).

But the very first step is still to find a host for such a repository.

--
----------------------------------------------------------------------------
Gerd Stolpmann      Telefon: +49 6151 997705 (privat)
Viktoriastr. 100             
64293 Darmstadt     EMail:   Gerd.Stolpmann@darmstadt.netsurf.de (privat)
Germany                     
----------------------------------------------------------------------------

Re: If i had a hammer...

[Markus Mottl]

> - Define a standard for documentation (very important for free software
>   development) -- we can simply use an existing one such as 'info'
>   or 'perldoc', or simply define a directory structure for html-pages

Good documentation is certainly a necessity!

> - Define a standard how to compile and install projects such that they
>   can be used together
> 
> For the latter, I think a meta-makefile mechanism would be appropriate.
> As Ocaml links always statically, there should be a tool
> 
> - that has as input a descriptive request what modules should be linked in
> 
> - that outputs a perfect Makefile that knows the linking order and automatically
>   calls ocamlc/ocamlopt/ocamlmktop -- whatever meets the given requests best

Although certainly not perfect (improvement suggestions welcome),
you might want to download my Makefiles which automate a great part
of the building process: it handles ocamllex- and ocamlyacc-files,
threads-library, automatic dependency calculation, building of byte code
and native code, as executable or as library.

Link: http://miss.wu-wien.ac.at/~mottl/ocaml_sources/intro.html

The distribution comes with an example.

[snip]
> read manuals very carefully to get the modules working. It should not be
> necessary to study and configure makefiles to do a simple thing like a
> software installation (people often think that installation is a simple
> process and a more or less an unproductive task).

What I really like about the distribution of OCAML and its tools: they
compile out-of-the-box (at least on all the machines I've tested).
Whoever has tried to compile or correctly install e.g. Mercury or (my
god!!) GHC, is certainly able to honour this...

Best regards,
Markus

-- 
Markus Mottl, mottl@miss.wu-wien.ac.at, http://miss.wu-wien.ac.at/~mottl

Re: If i had a hammer...

[Xavier Leroy]

I have followed this discussion with interest.  Having some sort of
central archive of publically-available Caml tools and libraries
would be an excellent thing, both to connect developers and users,
and to coordinate future developments.

I strongly encourage everyone who has written a piece of Caml code of
general interest to publish it on their web pages and announce it on
this mailing-list.

It would be great to have a set of Web pages listing all such
announcements.  We've been willing to do this for a long time, but
didn't quite find the time.  We could try harder, of course; but if
there are any volunteers for maintaining such a list (or other
Web material for Caml), we'd be very happy to provide an account on
the caml.inria.fr server.  (Please contact us directly at
caml-light@inria.fr.)

> I would really like to see a remote CVS-repository for OCAML somewhere. I
> am sure this would very strongly help people contribute collectively
> to projects written in OCAML. I have already asked at our university,
> but people there are a bit reluctant with software that they don't know
> so well and which allows remote access.
> Is there any machine at INRIA that could be used for CVS? With read-only
> access to the repository for everyone and write permissions for people
> who have already contributed useful code to a project?

We have plenty of old workstations that could be used for this
purpose, and indeed I've been thinking lately about providing
read-only CVS access to the OCaml development sources, as a simple way
to make available patches between releases.

Setting up such a machine raises delicate security issues (our
machines have been attacked twice in the last three years), and even
more so if read-write access is provided for some users.  Remote
developers would also need the ability to make .tar.gz distributions
of their sources available on our FTP server.  This can also be done,
but raises further security issues.

A more decentralized development model, where developers maintain
their own CVS archives and release on their own Web sites, would
certainly be easier to implement.  We would still have a centralized
listing of available software on our Web site, and perhaps automatic
mirroring on our FTP server (and on the INRIA Rocquencourt CD-ROM).

So, before we embark on setting up a public CVS server, I'd like to
know how many developers feel they could use it.  (Please respond
privately to caml-light@inria.fr.)

Let's start to fill that Caml's hump!

- Xavier Leroy

Re: If i had a hammer...

[Markus Mottl]

> 
> I have followed this discussion with interest.  Having some sort of
> central archive of publically-available Caml tools and libraries
> would be an excellent thing, both to connect developers and users,
> and to coordinate future developments.
> 
> I strongly encourage everyone who has written a piece of Caml code of
> general interest to publish it on their web pages and announce it on
> this mailing-list.
> 
> It would be great to have a set of Web pages listing all such
> announcements.  We've been willing to do this for a long time, but
> didn't quite find the time.  We could try harder, of course; but if
> there are any volunteers for maintaining such a list (or other
> Web material for Caml), we'd be very happy to provide an account on
> the caml.inria.fr server.  (Please contact us directly at
> caml-light@inria.fr.)

I guess that also maintainance can be decentralized to some extend,
though this raises security questions. I am not sure, how much "human
ressources" you have for supervising such a site. But it should be
possible to restrict "time consumption" of this task to things like
"granting (write) access", etc...

> We have plenty of old workstations that could be used for this
> purpose, and indeed I've been thinking lately about providing
> read-only CVS access to the OCaml development sources, as a simple way
> to make available patches between releases.

This would certainly be a great idea - this could speed up the development
cycle due to smaller "feedback cycles" from users.

> Setting up such a machine raises delicate security issues (our
> machines have been attacked twice in the last three years), and even
> more so if read-write access is provided for some users.  Remote
> developers would also need the ability to make .tar.gz distributions
> of their sources available on our FTP server.  This can also be done,
> but raises further security issues.

I see that this can be a problem. This strongly depends on the question,
in how far you can seperate such a machine from your "production network"
or any other security relevant facilities.

> A more decentralized development model, where developers maintain
> their own CVS archives and release on their own Web sites, would
> certainly be easier to implement.  We would still have a centralized
> listing of available software on our Web site, and perhaps automatic
> mirroring on our FTP server (and on the INRIA Rocquencourt CD-ROM).

I am not sure whether this would work so easily. Here some reasons:

 * many users do not have superuser rights on servers that are accessible
   "around the clock". The actual administrators might (as in my case)
   be reluctant with introducing remote repositories.
 * Administrative maintainance of such repositories would multiply -
   I am sure the overall effort is smaller if there is one central
   repository.
 * Some developers might not be this familiar with setting up such
   services.
 * Contributors would have to contribute to different repositories on
   different servers, which might be a bit confusing - especially if
   some projects "fit into the same category".

There are probably further arguments against this. I think it would be a
very logical approach to have such a site somewhere at INRIA. Of course,
any security questions would have to be solved first, but this certainly
depends on facts only INRIA can influence...

> Let's start to fill that Caml's hump!

So that it may traverse the deserts of software engineering!

Best regards,
Markus Mottl

-- 
Markus Mottl, mottl@miss.wu-wien.ac.at, http://miss.wu-wien.ac.at/~mottl