Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Alain Frisch]

On Mon, 14 Oct 2002, Stefano Zacchiroli wrote:

> I can't understand this recurring obstination: we _have_ a package
> manager, it's name is 'findlib'.
> It's ligthweight, simple, and do its dirty work well.

Of course findlib is useful !  But you can't do:

ocamlfind fetch some_package

to have findlib automatically download and install the last version of
some_package and its dependencies. I guess this is the kind of
fonctionality people mean when referring to a package manager.

-- Alain

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Stefano Zacchiroli]

On Mon, Oct 14, 2002 at 06:42:56PM +0200, Alain Frisch wrote:
> Of course findlib is useful !  But you can't do:
> 
> ocamlfind fetch some_package
> 
> to have findlib automatically download and install the last version of
> some_package and its dependencies. I guess this is the kind of
> fonctionality people mean when referring to a package manager.

Ok, anyway this feature can be easily implemented in the findlib
package, but a lot of new issues raise then.
Mainly:

1) does we really need such a feature, aren't distribution born for such
   issues? (obviously I'm thinking about the Linux world here)

2) if we really need such a feature (for example just to have
   transparent access to packages between different linux distributions
   or between linux and windows and mac and ... worlds) who is
   responsible to keep an archive of available package gettable via
   ocamlfind or whatevere else?
   CPAN, APT, and other approaches work because someon set up an
   official or de-facto-official archive (Well, APT also support other
   repositories, but official one are the most used and trusted ...).

3) ocaml packages are usually distributed as source packages and we have
   not a standardized way to build and install them, the success of
   CPAN, APT, are due to the facts that _or_ the distribution is in
   binary form with standardadized file system sructure _or_ the
   distribution is in source form with standardized compilation and
   installation procedure

Cheers.

-- 
Stefano Zacchiroli - undergraduate student of CS @ Univ. Bologna, Italy
zack@cs.unibo.it | ICQ# 33538863 | http://www.cs.unibo.it/~zacchiro
"I know you believe you understood what you think I said, but I am not
sure you realize that what you heard is not what I meant!" -- G.Romney
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Alain Frisch]

On Mon, 14 Oct 2002, Stefano Zacchiroli wrote:

> 1) does we really need such a feature, aren't distribution born for such
>    issues? (obviously I'm thinking about the Linux world here)

Well, Linux is not the center of the world (even of the Unix-world). There
are several OS, several distributions, and you're not always sys admin on
the system you're working on (or don't want to use root priveleges to
install an ocaml package).


> 2) if we really need such a feature (for example just to have
>    transparent access to packages between different linux distributions
>    or between linux and windows and mac and ... worlds) who is
>    responsible to keep an archive of available package gettable via
>    ocamlfind or whatevere else?
>    CPAN, APT, and other approaches work because someon set up an
>    official or de-facto-official archive (Well, APT also support other
>    repositories, but official one are the most used and trusted ...).

What about a Central OCaml Archive (with a nice acronym; or simply a
centralized list of URLs) and some conventions about namespaces ?

If it is impossible to manage such a central repository, an intermediate
solution would be to provide explicitly URLs for findlib-dependencies.

> 3) ocaml packages are usually distributed as source packages and we have
>    not a standardized way to build and install them, the success of
>    CPAN, APT, are due to the facts that _or_ the distribution is in
>    binary form with standardadized file system sructure _or_ the
>    distribution is in source form with standardized compilation and
>    installation procedure

A standardized compilation and installation procedure would be great,
indeed.

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Sven LUTHER]

On Mon, Oct 14, 2002 at 07:39:31PM +0200, Stefano Zacchiroli wrote:
> On Mon, Oct 14, 2002 at 06:42:56PM +0200, Alain Frisch wrote:
> > Of course findlib is useful !  But you can't do:
> > 
> > ocamlfind fetch some_package
> > 
> > to have findlib automatically download and install the last version of
> > some_package and its dependencies. I guess this is the kind of
> > fonctionality people mean when referring to a package manager.
> 
> Ok, anyway this feature can be easily implemented in the findlib
> package, but a lot of new issues raise then.
> Mainly:
> 
> 1) does we really need such a feature, aren't distribution born for such
>    issues? (obviously I'm thinking about the Linux world here)

Well, i guess the problem is essentially for the windows folk, and the
solaris/linux user who are used to rebuilding their stuff themselves.

Also suppor on rpm based distribs is not as mature as we have in debian,
i think.

> 2) if we really need such a feature (for example just to have
>    transparent access to packages between different linux distributions
>    or between linux and windows and mac and ... worlds) who is
>    responsible to keep an archive of available package gettable via
>    ocamlfind or whatevere else?
>    CPAN, APT, and other approaches work because someon set up an
>    official or de-facto-official archive (Well, APT also support other
>    repositories, but official one are the most used and trusted ...).

If this happens, maybe we can prevail on inria and the ocaml team to
make disk space available to us ?

> 3) ocaml packages are usually distributed as source packages and we have
>    not a standardized way to build and install them, the success of
>    CPAN, APT, are due to the facts that _or_ the distribution is in
>    binary form with standardadized file system sructure _or_ the
>    distribution is in source form with standardized compilation and
>    installation procedure

Yes, this is the problematic part.

Friendly,

Sven Luther
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Pierre Weis]

[...]
> > 2) if we really need such a feature (for example just to have
> >    transparent access to packages between different linux distributions
> >    or between linux and windows and mac and ... worlds) who is
> >    responsible to keep an archive of available package gettable via
> >    ocamlfind or whatevere else?
> >    CPAN, APT, and other approaches work because someon set up an
> >    official or de-facto-official archive (Well, APT also support other
> >    repositories, but official one are the most used and trusted ...).
> 
> If this happens, maybe we can prevail on inria and the ocaml team to
> make disk space available to us ?

Yes, indeed. There is no problem of disk space here. The problem is to
grant access in a secure and reliable way.

> > 3) ocaml packages are usually distributed as source packages and we have
> >    not a standardized way to build and install them, the success of
> >    CPAN, APT, are due to the facts that _or_ the distribution is in
> >    binary form with standardadized file system sructure _or_ the
> >    distribution is in source form with standardized compilation and
> >    installation procedure
> 
> Yes, this is the problematic part.

I think we can set up a standardized compilation and installation
procedure (meaning a standard Makefile for libraries, based on the
configuration set up by the original Ocaml installation). We can start
with the Caml team maintained software (the ``bazaar'') and encourage
people to do the same for their own software by storing (and
distributing) it on our server. This could be a lot of book-keeping
but we could probably manage it...

Pierre Weis

INRIA, Projet Cristal, Pierre.Weis@inria.fr, http://pauillac.inria.fr/~weis/


-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Sven LUTHER]

On Mon, Oct 14, 2002 at 10:38:36PM +0200, Pierre Weis wrote:
> [...]
> > > 2) if we really need such a feature (for example just to have
> > >    transparent access to packages between different linux distributions
> > >    or between linux and windows and mac and ... worlds) who is
> > >    responsible to keep an archive of available package gettable via
> > >    ocamlfind or whatevere else?
> > >    CPAN, APT, and other approaches work because someon set up an
> > >    official or de-facto-official archive (Well, APT also support other
> > >    repositories, but official one are the most used and trusted ...).
> > 
> > If this happens, maybe we can prevail on inria and the ocaml team to
> > make disk space available to us ?
> 
> Yes, indeed. There is no problem of disk space here. The problem is to
> grant access in a secure and reliable way.

Well, debian uses a system with upload queues, and signed package witha
strong web of trust between developpers who are allowed to upload
packages. Sure we also have ssh access on most debian boxes, but this is
not necessary for uploading.

I am sure a scheme could be found for this kindof distribution, with an
upload queue, where you could anonymously upload packages, which get
scanned for secure signatures and uploaded onto the server, or something
such. Maybe you could go without signatures even, since after all, there
is nothing critical and absolutely needing root access in the ocaml
packages.

> > > 3) ocaml packages are usually distributed as source packages and we have
> > >    not a standardized way to build and install them, the success of
> > >    CPAN, APT, are due to the facts that _or_ the distribution is in
> > >    binary form with standardadized file system sructure _or_ the
> > >    distribution is in source form with standardized compilation and
> > >    installation procedure
> > 
> > Yes, this is the problematic part.
> 
> I think we can set up a standardized compilation and installation
> procedure (meaning a standard Makefile for libraries, based on the
> configuration set up by the original Ocaml installation). We can start
> with the Caml team maintained software (the ``bazaar'') and encourage
> people to do the same for their own software by storing (and
> distributing) it on our server. This could be a lot of book-keeping
> but we could probably manage it...

Well, the problem is more in the dependency handling, and could be worse
if you maintain binary distribution, like you do for windows. But
anyway, i don't believe you could really afford to do binary
distribution for something else than windows or maybe macOSX.

It is very difficult to do this nicely without an integrated
distribution. Mmm, maybe it can be done, but would need lot of work, and
anyway, inria don't has the vast multi arch build farm that debian has,
so i don't know if it would be meaningfull.

Friendly,

Sven Luther
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Gerd Stolpmann]

Am 2002.10.14 23:07 schrieb(en) Sven LUTHER:
> On Mon, Oct 14, 2002 at 10:38:36PM +0200, Pierre Weis wrote:
> > [...]
> > > > 2) if we really need such a feature (for example just to have
> > > >    transparent access to packages between different linux distributions
> > > >    or between linux and windows and mac and ... worlds) who is
> > > >    responsible to keep an archive of available package gettable via
> > > >    ocamlfind or whatevere else?
> > > >    CPAN, APT, and other approaches work because someon set up an
> > > >    official or de-facto-official archive (Well, APT also support other
> > > >    repositories, but official one are the most used and trusted ...).
> > > 
> > > If this happens, maybe we can prevail on inria and the ocaml team to
> > > make disk space available to us ?
> > 
> > Yes, indeed. There is no problem of disk space here. The problem is to
> > grant access in a secure and reliable way.
> 
> Well, debian uses a system with upload queues, and signed package witha
> strong web of trust between developpers who are allowed to upload
> packages. Sure we also have ssh access on most debian boxes, but this is
> not necessary for uploading.
> 
> I am sure a scheme could be found for this kindof distribution, with an
> upload queue, where you could anonymously upload packages, which get
> scanned for secure signatures and uploaded onto the server, or something
> such. Maybe you could go without signatures even, since after all, there
> is nothing critical and absolutely needing root access in the ocaml
> packages.
> 
> > > > 3) ocaml packages are usually distributed as source packages and we have
> > > >    not a standardized way to build and install them, the success of
> > > >    CPAN, APT, are due to the facts that _or_ the distribution is in
> > > >    binary form with standardadized file system sructure _or_ the
> > > >    distribution is in source form with standardized compilation and
> > > >    installation procedure
> > > 
> > > Yes, this is the problematic part.
> > 
> > I think we can set up a standardized compilation and installation
> > procedure (meaning a standard Makefile for libraries, based on the
> > configuration set up by the original Ocaml installation). We can start
> > with the Caml team maintained software (the ``bazaar'') and encourage
> > people to do the same for their own software by storing (and
> > distributing) it on our server. This could be a lot of book-keeping
> > but we could probably manage it...
> 
> Well, the problem is more in the dependency handling, and could be worse
> if you maintain binary distribution, like you do for windows. But
> anyway, i don't believe you could really afford to do binary
> distribution for something else than windows or maybe macOSX.
> 
> It is very difficult to do this nicely without an integrated
> distribution. Mmm, maybe it can be done, but would need lot of work, and
> anyway, inria don't has the vast multi arch build farm that debian has,
> so i don't know if it would be meaningfull.

Just a few phenomenons that are hard to tackle :

- Sometimes a Makefile is not enough, you need a configure script. Sometimes
  the configure script takes arguments (Where can certain C libraries be found?
  Which build variant is chosen?). These arguments cannot be found out
  automatically, they are beyond the intelligence of the configure script.

  So an automatic package fetcher would have to deal with this complication.
  Maybe it has to ask interactively.

- Sometimes a single distribution tarball creates several libraries.

- Sometimes there are alternative dependencies (you can choose whether you
  like package X or X' as antecendent)

- Sometimes there are files to install that are neither libraries nor
  documentation. You need interaction with the operating system's idea
  of installing files, and if possible, with its package manager.

- Authors usually cannot test source packages on lots of platforms. Problems
  range from /bin/sh incompatibilities to the Unix/Windows nightmare. Especially
  for the latter I really don't know how to cope with it, because porting
  Makefiles and scripts to Windows is time-intensive, and as a Unix fan
  I have the attitude "let the Windows users do it".

These are not exoctic complications that are rarely found, they are normal.
A reasonable source distribution format must deal with them, and there must
be people willing to create such enhanced source packages. I don't say
this is all impossible, but sure it is an ambitious project.

Gerd
------------------------------------------------------------
Gerd Stolpmann * Viktoriastr. 45 * 64293 Darmstadt * Germany 
gerd@gerd-stolpmann.de          http://www.gerd-stolpmann.de
------------------------------------------------------------
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Sven LUTHER]

On Tue, Oct 15, 2002 at 12:44:57AM +0200, Gerd Stolpmann wrote:
> - Authors usually cannot test source packages on lots of platforms. Problems
>   range from /bin/sh incompatibilities to the Unix/Windows nightmare. Especially
>   for the latter I really don't know how to cope with it, because porting
>   Makefiles and scripts to Windows is time-intensive, and as a Unix fan
>   I have the attitude "let the Windows users do it".

At least for the windows solution, i strongly think a binary
distribution mechanism is what is needed for the users and also would
save a lot of time and 'howto build' kind of questions on the mailing
lists. There is no reason not to do so, i am no windows expert, but
windows offer a stable set of libraries and other system component that
can be relied upon to install ocaml binary packages upon. This already
works for ocaml itself.

> These are not exoctic complications that are rarely found, they are normal.
> A reasonable source distribution format must deal with them, and there must
> be people willing to create such enhanced source packages. I don't say
> this is all impossible, but sure it is an ambitious project.

Yes, i also think so.

Friendly,

Sven Luther
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Chris Hecker]

>At least for the windows solution, i strongly think a binary
>distribution mechanism is what is needed for the users and also would
>save a lot of time and 'howto build' kind of questions on the mailing
>lists. There is no reason not to do so, i am no windows expert, but
>windows offer a stable set of libraries and other system component that
>can be relied upon to install ocaml binary packages upon. This already
>works for ocaml itself.

Yes, binary distribution would work fine on windows.  Bytecode binary 
distribution for most libraries should work fine on all platforms.

I think at this point it's more important to do something small and get it 
going than it is to solve all possible package management problems.  Even 
something that just allowed a standard way of getting a tar ball of source 
for a given package and for its dependencies would be better than nothing, 
and we could improve it as we went.

The -pack option made this a lot simpler, because now libraries can be 
distributed as a single module.  If we make some relatively simple 
standards for build targets I think we can start making progress.

Of course, if something like apt would just work for us (on all platforms, 
source and binary), then that would be excellent and we should just start 
using it.

Chris

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Sven LUTHER]

On Mon, Oct 14, 2002 at 09:53:05PM -0700, Chris Hecker wrote:
> 
> >At least for the windows solution, i strongly think a binary
> >distribution mechanism is what is needed for the users and also would
> >save a lot of time and 'howto build' kind of questions on the mailing
> >lists. There is no reason not to do so, i am no windows expert, but
> >windows offer a stable set of libraries and other system component that
> >can be relied upon to install ocaml binary packages upon. This already
> >works for ocaml itself.
> 
> Yes, binary distribution would work fine on windows.  Bytecode binary 
> distribution for most libraries should work fine on all platforms.

Yes, that would be a good start, windows binaries and bytecode.

The outside dependencies (the ones on things not ocaml related, tcl/tk,
gtk+ and so on) is the main difficulty here.

BTW, if it comes to that, i have plan to ship every debian package
containing programs (as opposed to library packages) as a single
bytecode package built without -custom and a native code version on the
arches that support them.

If things like file layout, external dependencies and library
distribution (in particular the stublibs) are resolved, i suppose you
could install these exact same packages on non-debian systems also.

I don't have time to work on this though :(((

> I think at this point it's more important to do something small and get it 
> going than it is to solve all possible package management problems.  Even 
> something that just allowed a standard way of getting a tar ball of source 
> for a given package and for its dependencies would be better than nothing, 
> and we could improve it as we went.
> 
> The -pack option made this a lot simpler, because now libraries can be 
> distributed as a single module.  If we make some relatively simple 
> standards for build targets I think we can start making progress.

But the -pack option is not yet standardized, and not used in much
cases, it seems to me.

> Of course, if something like apt would just work for us (on all platforms, 
> source and binary), then that would be excellent and we should just start 
> using it.

it will never just work, it 'just works' on debian, because of the
debian maintainers who do the work of making sure the packages work.

Friendly,

Sven Luther
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Chris Hecker]

>Yes, that would be a good start, windows binaries and bytecode.

Yep, I think the right thing to do at the start would be to support windows 
binaries, bytecode for those libraries that don't need external C code 
libraries, and "opaque" source tarballs (by opaque, I mean the package 
manager will fetch the source code tarball just like it will fetch the 
binaries (and all the dependencies), but it won't attempt to build it or 
make it work).

This would get the right stuff onto your machine, which is a start.  Then 
we publish some guidelines and slowly the source becomes more and more 
consistent as people update, and then maybe we can solve the compilation 
problem.  But at least we'll have started.

>The outside dependencies (the ones on things not ocaml related, tcl/tk,
>gtk+ and so on) is the main difficulty here.

Yes.  We just punt that problem for now.

>But the -pack option is not yet standardized, and not used in much
>cases, it seems to me.

It's brand new, so give it time (and we need to get it working on windows, 
but that's close to solved now).  Is there a problem with it that you're 
implying, or just that it hasn't taken off yet?

> > Of course, if something like apt would just work for us (on all platforms,
> > source and binary), then that would be excellent and we should just start
> > using it.
>it will never just work, it 'just works' on debian, because of the
>debian maintainers who do the work of making sure the packages work.

No, I just meant could we use the apt executable and data formats and not 
have to write any code that fetches files or checks dependencies and 
whatnot?  In other words, would it save time to try to use it, or is it 
totally hard-coded to debian?

>I don't have time to work on this though :(((

Me neither, at least not for a year.

Chris


-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Sven LUTHER]

On Mon, Oct 14, 2002 at 10:25:51PM -0700, Chris Hecker wrote:
> >But the -pack option is not yet standardized, and not used in much
> >cases, it seems to me.
> 
> It's brand new, so give it time (and we need to get it working on windows, 

It was in since 3.04, wasn't it ?

> but that's close to solved now).  Is there a problem with it that you're 
> implying, or just that it hasn't taken off yet?

Just that it is not used yet. You know from previous threads that i
strongly advocate its use for all libraries, including pervasives, the
standard library. This is for namespace control, not distribution.

> >> Of course, if something like apt would just work for us (on all 
> >platforms,
> >> source and binary), then that would be excellent and we should just start
> >> using it.
> >it will never just work, it 'just works' on debian, because of the
> >debian maintainers who do the work of making sure the packages work.
> 
> No, I just meant could we use the apt executable and data formats and not 
> have to write any code that fetches files or checks dependencies and 
> whatnot?  In other words, would it save time to try to use it, or is it 
> totally hard-coded to debian?

Well, apple uses it (for darwin) so there should be no portability
issues. But then, i would have done a ocaml reimplementation, which
would allow to do some more advanced stuff, like checking for external
dependencies, which are not packaged. This cannot be done with the
standard dpkg/apt.

> >I don't have time to work on this though :(((
> 
> Me neither, at least not for a year.

Same here.

Friendly,

Sven Luther
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Chris Hecker]

> > >But the -pack option is not yet standardized, and not used in much
> > >cases, it seems to me.
> > It's brand new, so give it time (and we need to get it working on windows,
>It was in since 3.04, wasn't it ?

Nope, 3.05, so it's pretty new!

>You know from previous threads that i
>strongly advocate its use for all libraries, including pervasives, the
>standard library. This is for namespace control, not distribution.

I agree, hopefully the whole standard library will be -packed in the next 
release (assuming we fix the windows -pack, which is likely).  Perhaps that 
will break too much stuff, though...

There's also the question of how smart the linker is on packed 
objects...hopefully it will only link used modules and functions (and the C 
linker will do the same).  A few tests could answer this...

>Well, apple uses it (for darwin) so there should be no portability
>issues. But then, i would have done a ocaml reimplementation, which
>would allow to do some more advanced stuff, like checking for external
>dependencies, which are not packaged. This cannot be done with the
>standard dpkg/apt.

Yes, but if we don't even have time to use apt, we definitely don't have 
time to rewrite it.  :)  I think going with something that works but is 
nonoptimal is the way to start making progress on this problem.

Chris

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Alessandro Baretta]

Sven LUTHER wrote:
> On Tue, Oct 15, 2002 at 12:44:57AM +0200, Gerd Stolpmann wrote:
> 
>>- Authors usually cannot test source packages on lots of platforms. Problems
>>  range from /bin/sh incompatibilities to the Unix/Windows nightmare. Especially
>>  for the latter I really don't know how to cope with it, because porting
>>  Makefiles and scripts to Windows is time-intensive, and as a Unix fan
>>  I have the attitude "let the Windows users do it".
> 
> 
> At least for the windows solution, i strongly think a binary
> distribution mechanism is what is needed for the users and also would
> save a lot of time and 'howto build' kind of questions on the mailing
> lists. There is no reason not to do so, i am no windows expert, but
> windows offer a stable set of libraries and other system component that
> can be relied upon to install ocaml binary packages upon. This already
> works for ocaml itself.
> 

Very true, but I have a feeling most users/developers are on 
'nixes and don't really give a dime about Windows. I expect 
most people to be willing to release their working 
sources--working on their flavor of unix--but not to work on 
creating a Windows distro, be it source or binary. I also 
don't expect anyone, say, of the Debian folks, to pick up 
the Windows-port crusage single-handedly.

Let's just say this is a pretty nasty topic. I have a 
feeling that the only solution would be to have an official 
O'Caml IDE à la DevStudio--eeek!--capable of handling 
compilation and packaging in just the right way wherever 
it's used, without resorting to such unix tools as /bin/sh, 
configure, make and whatnot, but this would fit better in a 
commercial project than in an open source project.

The other solution is to start using Windows installation 
discs for what there best at: saving your tablecloth from 
beer stains...

Alex

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Chris Hecker]

>of handling compilation and packaging in just the right way wherever it's 
>used, without resorting to such unix tools as /bin/sh, configure, make and 
>whatnot, but this would fit better in a commercial project than in an open 
>source project.

I don't think it's a problem to require cygwin to be installed to build 
ocaml stuff, even the msvc build requires it now.  That's fine.  It's 
whether the resulting executables require cygwin (they shouldn't) or less 
importantly, whether they were built with msvc (they should be able to be 
in addition to cygwin/mingw).

Chris


-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Sven LUTHER]

On Tue, Oct 15, 2002 at 07:02:31AM +0200, Alessandro Baretta wrote:
> 
> 
> Sven LUTHER wrote:
> >On Tue, Oct 15, 2002 at 12:44:57AM +0200, Gerd Stolpmann wrote:
> >
> >>- Authors usually cannot test source packages on lots of platforms. 
> >>Problems
> >> range from /bin/sh incompatibilities to the Unix/Windows nightmare. 
> >> Especially
> >> for the latter I really don't know how to cope with it, because porting
> >> Makefiles and scripts to Windows is time-intensive, and as a Unix fan
> >> I have the attitude "let the Windows users do it".
> >
> >
> >At least for the windows solution, i strongly think a binary
> >distribution mechanism is what is needed for the users and also would
> >save a lot of time and 'howto build' kind of questions on the mailing
> >lists. There is no reason not to do so, i am no windows expert, but
> >windows offer a stable set of libraries and other system component that
> >can be relied upon to install ocaml binary packages upon. This already
> >works for ocaml itself.
> >
> 
> Very true, but I have a feeling most users/developers are on 
> 'nixes and don't really give a dime about Windows. I expect 
> most people to be willing to release their working 
> sources--working on their flavor of unix--but not to work on 
> creating a Windows distro, be it source or binary. I also 
> don't expect anyone, say, of the Debian folks, to pick up 
> the Windows-port crusage single-handedly.

A, no, i don't think we are going to be willing to do this, altough to
be honest, i spend very little time under windows, well almost none, but
when i go there to check if my program also run under windows, i very
much like to have everything directly installable and working, without
having to worry how to build stuff and such.

So there is an interest here.

Disclaimer: in this, i am speaking solely for me, not for the rest of
the debian/ocaml team.

> Let's just say this is a pretty nasty topic. I have a 
> feeling that the only solution would be to have an official 
> O'Caml IDE à la DevStudio--eeek!--capable of handling 
> compilation and packaging in just the right way wherever 
> it's used, without resorting to such unix tools as /bin/sh, 
> configure, make and whatnot, but this would fit better in a 
> commercial project than in an open source project.

No, the right way is to make a binary distribution, with someone (inria
or the consortium) building the libraries and ocaml and making it easily
available. This would maybe need someone devoting its time to port
widely used non inria software, or nag the authors of it to fix windows
build whenever possible.

Friendly,

Sven Luther


-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] CDK with Ocaml 3.06 (fwd)

[Gleb N. Semenov]

Gerd Stolpmann wrote:
> 
> Am 2002.10.14 23:07 schrieb(en) Sven LUTHER:
> > On Mon, Oct 14, 2002 at 10:38:36PM +0200, Pierre Weis wrote:
> > > [...]

[...]

> > It is very difficult to do this nicely without an integrated
> > distribution. Mmm, maybe it can be done, but would need lot of work, and
> > anyway, inria don't has the vast multi arch build farm that debian has,
> > so i don't know if it would be meaningfull.
> 
> Just a few phenomenons that are hard to tackle :
> 
[...]
> 
> These are not exoctic complications that are rarely found, they are normal.
> A reasonable source distribution format must deal with them, and there must
> be people willing to create such enhanced source packages. I don't say
> this is all impossible, but sure it is an ambitious project.
> 
> Gerd

Little comment.

Yes, all said is almost true. But perl people had solved the same
problem
with Makefile.PL included in any source distributuon of perl modules,
hadn't them?


Regards!
GNS

-- 
Gleb N. Semenov		111621, Muromskaya St. 21, apt. 2, Moscow, Russia 
gleb@ahome.ru        	phone +7(095)700.0172
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Sign the packages (was Re: [Caml-list] CDK with Ocaml 3.06 (fwd))

[Tim Freeman]

>...anonymously upload packages...
>Maybe you could go without signatures even, since after all, there
>is nothing critical and absolutely needing root access in the ocaml
>packages.

This is quite dangerous.  Anyone can then anonymously upload a new
version of any package that starts by doing "rm -rf ~".

-- 
Tim Freeman       
tim@fungible.com
GPG public key fingerprint ECDF 46F8 3B80 BB9E 575D  7180 76DF FE00 34B1 5C78 
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: Sign the packages (was Re: [Caml-list] CDK with Ocaml 3.06 (fwd))

[Sven Luther]

On Tue, Oct 15, 2002 at 04:21:03AM -0700, Tim Freeman wrote:
> >...anonymously upload packages...
> >Maybe you could go without signatures even, since after all, there
> >is nothing critical and absolutely needing root access in the ocaml
> >packages.
> 
> This is quite dangerous.  Anyone can then anonymously upload a new
> version of any package that starts by doing "rm -rf ~".

Well, sure. So you have 2 choices :

  o require signed uploads from people you trust (and can visit
    consequence upon if they misbehave)

  o doing it so it will not be installed as root.

Mmm, i did not see that it would erase the home directory. Well, this is
only possible if the packages include scripts to be run before or after
the install. This is what debian uses right now, but then we use
solution 1 above.

Friendly,

Sven Luther
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners