Actually this idea could work for our de-centralized pardigm, imagine that
each server only has knowledge of 6 other server, and each of those know 5 different
machine. Now calculate this, in this case if we decide to only travarse down
this tree 6 times
we can end up with look up sizes at the size of (6^5)+6 machines ( which ends
up about beign
7781 look ups) , easily this is assuming that each entry is unique.
I'll write up more about it, but what do you think?
Also I think we should use a Resource Locator String to indentify services on
other machines
this means that a user could use a single login string for all services on
DotGNU
because the string points back to a server which has the authorization info.
Good idea :)
I like the idea of having the server check as well, however, if there
are severe network problems, this could only make things worse.
(Network storm anyone?)
As such, I suggest we set server checks to be done once in a certain
amount of time for the entire server, once that time slot expires, a new
check occurs. This way, 2000 clients aren't requesting server uptime
checks from a single server all at once :)