Re: Central vs. Distributed Archives

From: Greg Kuperberg <greg_at_MATH.UCDAVIS.EDU>
Date: Wed, 8 Nov 2000 13:48:40 -0800

On Wed, Nov 08, 2000 at 12:30:39PM -0400, David Goodman wrote:
> Departments are not the place, for exactly the reasons John explains. More
> than one of the academic depts. in more than one major university I have been
> affiliated with has managed to lose unique copies of Ph.D. theses, as well as
> every other possible type of item.

The fact is that most math papers on the web (excluding those in the
arXiv) are on department, and not campus-wide, web servers. This is
even true of papers that are organized into preprint series. One of the
dangers of an interoperability approach is to hoist the e-print vision
on such an accidental foundation. I also agree with John MacColl's
position that libraries are more reliable archivists than departments
in principle. But I disagree entirely with the claim that distributed
interoperability has never been tried before. It has been tried several
times, whole-heartedly with these two projects:


And it has been a factor in many other projects, including Hypatia
and the AMS preprint server. Some of these projects are more
successful than others, but *all* of them suffer from inconstancy
of the underlying archives.

While libraries certainly should help preserve e-prints, I do not trust
any one library, nor any other sole institution, to archive material
single-handedly. Any caretaker can lose or destroy a unique copy of
any document. (Just last year the Boston Public Library lost thousands
of books in a flood, for example.) That is why it is important to
redundantly and openly mirror an archive and not just allow third-party
searches. The arXiv has 18 mirror sites on six continents, listed at:

That is not as many copies of the arXiv as I would like to see, although
it is enough full-fledged active mirrors. More significantly anyone
who wants to can maintain yet another copy of the arXiv following the
instructions at:

As a rule, it is better for web sites to share the same archive than
to each have fragments. It is better for Oxford and Cambridge to
each have all of Shakespeare's plays than for Oxford to have only the
comedies and Cambridge to have only the tragedies. That is why I favor
shared interoperability, which is in some ways centralized, to fragmented
interoperability, which is optimistically called decentralized. Massive
redundancy is one of the few strengths of the existing paper-based system;
let's not tear up the road in addition to scrapping the horse carriage.
  /\  Greg Kuperberg (UC Davis)
 /  \
 \  / Visit the Math ArXiv Front at
  \/  * All the math that's fit to e-print *
Received on Mon Jan 24 2000 - 19:17:43 GMT

This archive was generated by hypermail 2.3.0 : Fri Dec 10 2010 - 19:45:57 GMT