Much as I still have forrest issues with Python Gump, and some subtle
working differences/failures, I think the core of Python Gump is there. I
believe that the build on LSD is sound, just suffering from CVS servers
issues, and again -- a few subtle differences that I can't put my finger on
this moment. Since traditional gump was insensitive to CVS problems (I
believe it failed to react to them) and Python Gump is ultra sensitive, I
think much/most of the issues can be attributed to this.
See: http://lsd.student.utwente.nl/gump/

I've made progress w/ Nick's "packages overriding real modules/projects" (I
was incorrectly checking dependencies on packages) but I've more to do.
See: http://gump.chalko.com/py/

As such, while I will keep hacking at bugs as I find them, I'd like to get
Gump (and Sam/others) out of the manual package installation business...

I'd like to discuss:

* Gump calling Ruper (to install packages)
* Gump producing a JAR Repository (for other subsidiary Gumps to feed off)

I'll put some thoughts here, and if others have other/conflicting/additional
thoughts maybe we could work on the Wiki. I am open minded, and game to
implement pretty much anything. Anyways, my thoughts...

* Gump calling Ruper:

If we add an entry into a project [or module] descriptor (that is packaged)
along the lines of:

    <repository
                id="maven"
                group="eclipse" />

.. and no CVS and no <ant|<script tags, we could have Gump use Ruper to
download/synchronize the latest version at that repository with the version
in the local package tree. Effectively this would keep a *single copy* of
the *latest* of some package on the gump. This seems a nice balance, one
only copy of non-Gumped projects, and the latest -- to keep folks for going
stale.

Other thoughts I have are along the lines of <project depend="x"
version="1.2.3-alpha" combined with the above, to have folks be able to pick
their version. This is doable [I believe], but more work and less
"gumplike".

*Gump producing a JAR Repository

Along those lines, I am thinking that if Gump produced a "named hierarchical
repository" (along the lines of what Jakarta infrastructure agree upon, I
believe) we could have subsidiary gumps get "the latest package" from one or
more upstream Gumps. I think this would make the lives of Nick/I (and other
who run local Gumps) much easier, and save the OSS world a bunch of cycles.
I know there is a can of worms with this, but I like the concept of a tree
of Gumps all working together efficiently. I think we need more and more OSS
projects to Gump, and I don't think that any one Gump site has resources for
all that, but a tree might have.

regards,

Adam
--
Experience Sybase Technology...
http://www.try.sybase.com


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to