On Aug 10, 2004, at 11:10 AM, Andreas Raab wrote:

>> "Hey! Where did you find that? Oh, I didn't know about *that* 
>> server....
>> Hmmm, it isn't up now, do you have a copy you can email me?"
>
> The concern here is that having multiple universes reduces the 
> availability
> of packages; however (case in point) if your one central server is down
> you're dead in the water. What I find ironic about it is that what 
> we've
> seen today is the exact opposite of the above concern (the central 
> server
> down, the decentralized structure coming to the rescue) and that 
> strikes me
> as very ironic indeed considering that all of this happened within 
> about 24
> hours ;-)

It's also important to consider the distinction between a single map 
available from multiple servers and multiple maps.

If SqueakMap were mirrored, failure of a single server would make no 
difference - SM clients could fall back to the mirrors. In a more 
distributed system where you have multiple maps, it's possible to have 
partial outages, where most of the the map is available, but the 
packages from a given server are not. Yes, it's certainly possible set 
up mirrors to avoid partial outages, but I suspect this is less likely 
to happen in the "more distributed" scenario - probably only for big, 
important servers.

As far as I can tell, Göran has never been against multiple servers 
(that is to say, mirrors for SqueakMap), but is vehemently against 
multiple maps.

Personally, I'm not arguing either way here, just pointing out that 
this robustness argument doesn't really apply to the question of 
whether or not the *map* should be centralized.

Colin