[Bug-gne]various random thoughts

[Rob Scott]

hello.
I'm back.

(*groan*)

exams over, so now i think we can get some stuff really done.
ive heard that arpith is back in action, so we should be able to
get a good team together to sort some things out.

i suppose the thing most on peoples minds right now is searching systems
etc, because this can really make or break a resource.

personally ive been putting some serious thought into mirroring
of the db's etc.
this is going to be difficult.
not only must the db be mirrored but the other resources too.
ie- images etc.

speaking of images, actually, we need to get them working.
I know its not possible to get them fully working without the server setup,
but, tom, you could use the various sourceforge servers to almost simulate 
this.
could you?
i dunno.

back on the subject of mirroring, we could use standard mysql replication,
however this would not cope with images.
on the other hand, I could write a simple client server app in C / sockets
to handle this.
however, if mysql replication is already available, whats the point of 
rewriting things already done?

mysql replication uses a thing called a binary log, which basically logs all
changes to a db, including deletion.
i could intercept this log to do things myself, but in a way it would be 
unnecessary work.

the problem with replication is that it requires a constant connection.
as soon as a change is made, it is cascaded down.
the advantage to rewriting this (or more accurately: redirecting this) 
ourselves is that it could be a proper client server thing.

the server would run on the master server and wait for a client to ask for 
updates, at which
point it would send all updates.

this needs far more thought on my part, clearly, but i thought i would 
share some of my ideas
incase anyone could suggest something.

i am still in favour of only being able to add articles to the master server.