Cobol is the new language to know?

Em 22-02-2011 05:12, Geert Claes escreveu:

>
> csrabak wrote:
>>> Focusing in  the squeak/pharo/cuis  branch, I've noticed  that pharo
>>> people got really concerned in
>>> aspects that  can attract  market interest.   They're looking for
>>> funding, fighting  to have "hired
>>> people" (meaning under wages) minding "hairy  aspects" of development,
>>> maintenance, documentation
>>> and some sort of "standardization", etc. Collaborators also produced
>>> interesting "printed material" (books,
>>> tutorials) but IMO such material is still very academic in nature. People
>>> in industry likes best "manual"
>>> stuff  (like "foundation  classes" with  "methods" (messages  in the
>>> case) documented,  etc). IMO (again)
>>> here lies a  relevant problem: there  are "no foundation  classes" (no
>>> "landmarks" or things that cannot  be  > easily changed) and it becomes
>>> apparent when people exchange ideas  about "keeping this"  and "getting
>>> rid of that". At least debate is going on and many issues have been
>>> addressed.
>>
>> Yes.
>>
> I don't understand what was said here?  Can someone elaborate?

In order to increase acceptance of smalltalk in corporate world (and
even in academic and scientific worlds) someone (pharo/squeak/cuis
people) must come out with a solution that:

1) Is well documented
2) Is supported
3) Is stable

To be well documented (in the sense corporate world understands it) a
smalltalk distro must:

1) Have a body (a set of) classes that are considered "foundation
classes". Those must be present in all versions of the distribution and
must have an unchanging set of features (in the case of smalltalk: an
unchanging set of messages,considering instance and class variables are
available through assessors). Those foundation classes must cover a
range of functionalities that makes the distro useful.
2) These foundation classes must be documented in a way that resembles,
at least, man pages of *nix (syntax, semantics, dependencies, known
bugs, exceptions, etc)
3) Foundation classes should be arranged in "orthogonal" packages
(meaning: if I unload one package everything that doesn't depend on it
still works). Anyways, dependencies should be documented in a very clear
and explicit way.

Just to give an example on how lack of documentation hurts open
smalltalk, how many people gave it up for not  being able to deliver
acceptable interfaces because Morphic is hideously documented? I know
several ones. In real corporate world, learning through trial and error
is not acceptable, just like blowing schedule up due to the need to
figure out by experience how ProportionalLayout works neither...

To be supported there must be people in charge of:

1) Debugging and fixing it (inside a well known schedule & following an
agreed maintenance protocol)
2) Elaborating suitable documentation (as described above)
3) Preparing didactic material (courses, books, etc)
4) Disseminating information of what's going on inside distro's
development (much in the way Ubuntu/Fedora/etc communities do)

Stable is stable...

There are other things that are important but IMO they all come after
ones I just wrote about.

Em 22-02-2011 14:58, Eliot Miranda escreveu:
>
>
> (...)

>
> IMO, while important,  this misses the most important features which
> together provide a useful platform:
>
> 1. excellent FFI (including threading support) so that the system can
> be integrated with other code, both as a user of libraries and as a
> provider of libraries
> 2. support for foreign files in Monticello so that Monticello can be
> used to manage complete projects, not just the Smalltalk assets
> 3. scripting and excellent file system facilities so that one can
> easily interact with the file system, still a vital part of the
> majority of software systems (the Squeak file system code is a cruel
> joke, still resembling the original Smalltalk-80 code from the late 70's).

+1 here. (1) and (3) are really important. I'd add enhancing network
streams too.
One thing that's overseen but is extremely important for developing
commercial software is full internationalization support (like i18n) and
its a major task since large sections of "foundation classes"
(TimeAndDate for instance) would have to be adapted/changed.

> Without the necessary functionality people simply can't use Squeak to
> produce their applications and are forced to choose something else.
>  These features, along with good web support, are what I understand by
> "platform".  The web support is also pretty good, with Seaside, SSL,
> ODBC, HTTP connectivity, etc being pretty good.  The community has
> made excellent progress over recent years.  The "well documented",
> "supported" and stable" features are not terrible.  Yes they can
> always be improved but on the whole that's not the fundamental block
> to effective use of Squeak/Pharo.
>
> Performance is also not an issue to getting started; plenty of Ruby
> and PHP projects sow that one can tackle performance later on.

Performance is a relative thing. As far as there are no limitations to
performance imposed by VM (and squeakvm/cogvm are really generous),
things can be continually improved.

There's one aspect of performance that can be an important show stopper
for pharo/squeak/cog: memory usage ceil at 4GBytes.

> Fundamental is providing a platform and we're not there yet.
>
> 2¢
> Eliot
>
>