On 3/30/11 6:57 PM, sebb wrote:
> On 31 March 2011 01:38, Phil Steitz <phil.ste...@gmail.com> wrote:
>> On 3/30/11 4:22 PM, Jochen Wiedmann wrote:
>>> On Tue, Mar 29, 2011 at 5:52 PM, Phil Steitz <phil.ste...@gmail.com> wrote:
>>>
>>>> I disagree with this.  The most important artifacts are the
>>>> zips/tars that go to dist/.  These *are* the ASF release.  Nexus
>>>> makes it *harder* IMO to maintain provenance of these artifacts.
>>> These artifacts are present in Nexus. Pulling them to a temporary
>>> directory is quite easy with wget.
>> And then you need to check the hashes and sigs again since you are
>> now working with downloaded copies of the files that we voted on.
> Huh?
>
> If we create a script to move the files directly from Nexus staging to
> dist/, surely there's no chance that the cp+rm will somehow mangle the
> files?
mv is no problem.  wget via http means you get a copy with the
network between.  That is what I was referring to.  In that case,
just like we tell the users, the RM should check hashes and sigs to
make sure that what actually ends up in dist/ is identical to what
we voted on.
>> Seems much easier and more correct to me to just scp the files to
>> p.a.o., let people vote on them and *move* exactly those files to
>> /dist.
> Which is much what happens with Nexus now, apart from the dist/ move
> phase which is not yet automated.
>
So the nexus staging repo lives on p.a.o?   Does not look like it
from the IPs, unless there is some aliasing going on.  If dist/ is
itself mirrored on r.a.o, then mv is possible; otherwise the files
have to be copied across the network, rather than moved.  That
requires a recheck of the hashes.

Phil
>>>  At which point I can see no
>>> difference between your proposed solution and this one. And there's
>>> nothing to do for all the other files that live in Nexus (and must
>>> live, because Maven is just too important, whether we like it or not).
>> Sorry, I don't buy that.  The tars and zips need to "live" in
>> /dist.  The maven artifacts need to make their way to the maven
>> mirrors.  Having a "staging" repo where we can inspect the maven
>> bits before they get pushed out is great (though I think our homes
>> on p.a.o are fine for this).  Why can't we just push files directly
>> there using scp or Ant tasks and then "promote" them to the rsynch
>> repo using a little script including commands like those in step 3
>> of http://commons.apache.org/releases/release.html?
>>>> I also don't see why we *must* rely on proprietary software to
>>>> manage replication.
>>> I'm mostly with you on that. I strongly opposed choosing Nexus in
>>> favour of Archiva for that very reason. But we have Nexus now and I
>>> wouldn't want to have Commons a swimmer against the rest of the Apache
>>> tide.
>> Based on Mark's response, I don't think we are the only ones :)
>>
>> Phil
>>> Jochen
>>>
>>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org
>> For additional commands, e-mail: dev-h...@commons.apache.org
>>
>>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org
> For additional commands, e-mail: dev-h...@commons.apache.org
>
>


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

Reply via email to