On Thu, Mar 3, 2011 at 4:37 AM, Mattmann, Chris A (388J) <[email protected]> wrote: > On Mar 2, 2011, at 8:56 PM, David Crossley wrote: > >> Mattmann, Chris A (388J) wrote: >>> Hmmm, >>> >>> I'd like to question this: why would the Incubator ever delete archives of >>> releases? We have active links on the TLP for e.g., OODT and for e.g., Tika >>> that point users to our Incubator releases that are still active Incubator >>> links. I realize that archive.apache.org keeps a copy, but following your >>> logic, why would any TLP maintain its own dist area? It could just point to >>> the archives? >> >> We are not deleting any archives. > > You're right, you're suggesting that it's good that we delete the original > copy, the one that release announcements link to via email and the ones that > folks may have permlinked as bookmarks, and the one that our own mail > archives link to? >
"dist" is a distribution area - mirrored so that when people download they usually end up on a mirror site. This spreads the load of people downloading our software. Projects shouldn't be linking to specific artifacts in the dist area - they should be linking to a download page where links to mirrors are available. If every project linked directly to actual artifacts in the dist area of the apache site then the whole mirroring to reduce the load on our infrastructure is subverted. The policy is to only keep the latest release in that *distribution* area so that it doesn't grow too large. Any artifact that is put there is automatically copied to the archive area. Those are the real archives, not older release left hanging around in the *distribution* area. See http://www.apache.org/dev/mirrors.html Niall >> Keeping the dist area tidy with only current releases >> means that this huge ASF-wide content is mirrored efficiently. > > At what cost? In my mind, there is a subtle point lost here. > > Archiving is good and it has to do with mirroring. What i'm advocating for is > preservation. We should never delete the original copy dists for *any > project*. It creates disconnects between the world-wide mirroring system that > is called email, and locally caching things, and archiving and cataloging > links. > > The cost of invalidating those negates a ton of the savings that you're > proposing. > >> >> Also, we don't mirror archive.a.o via that dist system. > > Gotcha. > >> >> I just had a look at Tika download page. It does as i suggested >> earlier: They link to >> archive.apache.org/dist/tika/ >> for old releases >> and to >> archive.apache.org/dist/incubator/tika/ >> for the older incubator releases. > > Yeah I think we updated that. I couldn't remember if we actually directly > linked to incubator.a.o. I think I might have even updated that when I > noticed the links were broke on the site. > > Regardless, we should haven't to update it it. I'm proposing that's *it's a > good thing* to have an original + 1 copy from a reliability and software > preservation perspective. > > Cheers, > Chris > > ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ > Chris Mattmann, Ph.D. > Senior Computer Scientist > NASA Jet Propulsion Laboratory Pasadena, CA 91109 USA > Office: 171-266B, Mailstop: 171-246 > Email: [email protected] > WWW: http://sunset.usc.edu/~mattmann/ > ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ > Adjunct Assistant Professor, Computer Science Department > University of Southern California, Los Angeles, CA 90089 USA > ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > > --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
