[AG-TECH] TVS Upgrade Wednesday

[Markus Buchhorn]

At 19:43 10/03/2004 -0600, Thomas D. Uram wrote:
>- The venue server is now running with a monitor alongside.  The monitor periodically tests the server accessibility, and restarts it if it becomes unavailable.  This is a win over the previous behavior, where it had to be restarted manually and could be unavailable for a longer period.

I vaguely recall an article about 12-15 years ago, where somebody noted that "Unix-robustness" was "Write it, if it runs for a while then great. Then write a boss-server to sit above it and restart it *when* it crashes". :-) 

Anyway, vague question ahead :-)

Have you considered getting venue servers to communicate amongst each other, building a DNS-or-p2p-like system of redundant venue servers? They could share information on 'public' venues, and still host a set of 'local/private' venues. Then, when one goes down, the venue client can query a remote/cached list of alternate venue servers. 

As far as addressing goes, this should be easy. The concept of stored documents/objects makes it trickier, although again if they were replaced with references rather than the actual 5GB DoS objects that some people ship round ;-) that could be managed as well.

The additional benefit is that the nuisance of "venue at venueserver" can be cleaned up a bit. I can still address venues that way, but today there's two different UI paths within the current venue client (pick venueserver via menu/URL, pick venue via N_layers*(door list) ) when it could be one.

Cheers,
        Markus


Markus Buchhorn, ANU Internet Futures Group,       |Ph: +61 2 61258810
Markus.Buchhorn at anu.edu.au, mail: Bldg #108 CS&IT  |Fx: +61 2 61259805
Australian National University, Canberra 0200 Aust.|Mob: 0417 281429