Re: Who keeps a (copy of) that old wrench? [REPEAT]

[Barry McMullin]

>>Date: Mon, 01 Jul 1996 08:27:00 -0700
>>From: Sven Thommesen <address@hidden>
>>Subject: missing mail
>>
>>I just discovered that there's a couple of messages at the end of the 
>>Swarm support archive for 9606 (from Friday the 28th) that never made it 
>>to the list (at least, to 3 locations I can verify). They are the 
>>messages by Barry McMullin and JJ Merelo on the subject of "Swarm 
>>Archives". 
>>
>>Perhaps someone could retransmit those? The net must have been overloaded 
>>this weekend ...

[Okey Dokey: here's my message again anyway: JJ will have to speak
for himself. Apologies to everyone who's already seen it. - Barry.]

At 11:39 27/06/96 -0700, Sven Thommesen <address@hidden> wrote:
[...]
>I think perhaps there is a larger question here for the whole Swarm
>community to think about:
[...]
>If we as a community are serious about making replication of
>our experiments possible, then by the nature of the beast old versions
>of Swarm, tcl, tk, BLT, libtclobjc etc. must be saved along with every
>Swarm simulation, *or* they must be available from some repository.
[...]
>Comments welcome.

I'm just starting to use swarm (finally got heatbugs running last
Tuesday!) - but I've also been mulling over this problem.  This may
be a crazy idea, but I'm not sure its realistically avoidable:
the solution may be not just to archive the *software*, but to
"archive" the whole *computer*!

The problem is that the mutual dependencies don't stop somewhere 
cleanly - either within the software, or even at the hardware/software
interface.  If we are *serious* about promising long term replicability,
I think that means keeping the whole original "instrument" available.

(I use the word "instrument" advisedly here: one could make a rough
analogy with something like a large telescope serving the astronomy
community.)

How would this possibly work?  I suggest that institutions (including,
but not limited to SFI) could make available public, or network
accessible, swarm platform(s).  While all of us could, and should,
continue to do swarm development on private facilities, we would
have to generate published results on one of the "public" platforms.
Anyone wishing to replicate or criticise the published results would
get access to this public platform to work on (rather than having to
try to rebuild a compatible configuration from scratch).

There are no very serious *technical* difficulties with this - as
Internet bandwidth expands, running X connections from anywhere in
the world becomes perfectly realistic.

But there *are* significant resource questions, of course (more
serious than just archiving the "software"): institutions would have
to commit to maintaining "obsolete" swarm platforms publically
online, in their obsolete configuration, for a fairly indefinite
period...

Of course, it may be that useful swarm work will be completed on
relatively modest platforms (a high end linux PC, costing less
than maybe US$5000): in such cases I think it perfectly feasible that
that specific machine be *frozen*, and left on the net, at the
conclusion (publication) of the work...

Hmmm.  Maybe it *is* totally crazy...

Barry.

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
| Barry McMullin, Autonomous Systems Group,  |    address@hidden |
| School of Electronic Engineering,          |  Voice: +353-1-704-5432 |
| Dublin City University, Dublin 9, IRELAND. |  FAX:   +353-1-704-5508 |
| http://www.eeng.dcu.ie/~mcmullin/home.html |                         |
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++