commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mark R. Diggory" <mdigg...@latte.harvard.edu>
Subject Re: [collections] new snapshot to ibiblio
Date Fri, 14 May 2004 01:39:23 GMT
Outlining our current goals concerning snapshots:

1.) The establishment of the "Apache Repository" was a two-fold purpose.

(a) Distributions of Apache jars were not actually being maintained at 
Apache but ending up maintained at Ibiblio, this is in conflict with 
Apaches Release and Mirroring Procedures.

(b) Apache developers actually had to go through a request process to 
get jars onto what amounts to a "mirror" of Apache content. This is 
backward in comparison to Apache being the canonical source for all mirrors.

So Jason and I resolved this by establishing the Apache Repository for 
Apache developers and having it rsynced by ibiblio. Thus Apache is again 
the canonical source for all Apache content. I currently maintain the 
rsync via my ibiblio account.


2.) Because we established the repository under 
/www/www.apache.org/dist/java-repository, it is also mirrored to all 
other Apache mirrors.

(a) The eventual goal is to actually use this process to distribute 
download requests for jars to the mirrors and distribute the repository 
contents more uniformly across the internet, reducing any load on Apache 
servers, this is a rather dispersed effort across various projects like 
Maven and Depot.

(b) Because /dist/ is mirrored, it is reserved ONLY for official signed 
releases and has strict guidelines pertaining to this. However, some 
jars remain in the Apache Repository which are not official releases 
strictly for historical reasons.

We are highly discouraging the release of any dated build snapshots into 
the Apache Repository because of these issues, plus the fact that we 
would rather have external projects dependent on stable releases as 
opposed to unstable nightly or dated builds.

To resolve the need for internal developers to be dependent on latest 
unstable builds of a project, we setup the 
/www/cvs.apache.org/repository to support those dated builds.

There is a convergence between the Gump Integration and this unstable 
repository which I hope is occurring, eventually, we hope that Gump will 
be used to produce a nightly build repository for projects, and that 
most projects will start taking advantage of this Continuous Integration 
for their development. I have been watching (lurking?) the Gump folks 
diligently working on this and I think its coming along.

Until this is complete, the unstable repository is available for those 
who need to release unstable builds. Because it is on cvs.apache.org, it 
is not mirrored, nor does it endup at ibiblio. This unstable repository 
should not be utilized by non-apache developers, eventually, we may 
consider some form of security to restrict access to it.

I hope this clarifies our efforts so far to date. I understand there are 
still some kninks in the system. But with time and effort, we will 
eventually have a streamlined distributed repository system going.

-Mark

Stephen Colebourne wrote:

> I have recently had to ask for a rogue collections snapshot to be changed to
> point at the 3.0 release. This had been creating a negative impression of
> collections.
> 
> Consider this just my opinion, that it would be nice to have snapshots
> clearly separate from releases (two directories in the repository for
> example).
> 
> In this particular case, I am happy for Mario to put a collections snapshot
> in the Apache repository, but hopefully collections 3.1 will be along soon
> anyway.
> 
> Stephen
> 
> From: "Dain Sundstrom" <dain@coredevelopers.net>
> 
>>On May 13, 2004, at 2:23 PM, Stephen Colebourne wrote:
>>
>>>I am opposed to adding snapshots to ibiblio, as I have seen it create
>>>isues.
>>>IMHO ibiblio should be released/stable code only.
>>
>>Can you be more clear?  I think ibiblio snapshots are great and would
>>hate to see them go away.  Think of all the projects out there that are
>>using apache snapshots that would have to add the apache repo to their
>>project.  Not only is this a lot of traffic to apache, I think it also
>>sets a bad precedent for other projects.  Think of a project using a
>>lot of snapshots and they have to add every small project's repo to the
>>repo list (which may also add to the apache traffic as maven has no
>>idea which repo may have the snapshot, so it tries them all in order)
>>Anyway I'm rambling now....  I am curious about the issues this
>>creates.
> 
> 
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: commons-dev-help@jakarta.apache.org
> 

-- 
Mark Diggory
Software Developer
Harvard MIT Data Center
http://www.hmdc.harvard.edu

---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org


Mime
View raw message