Regarding tcpserver/tcpclient these don't cut it as they only forward numerical 
info. We need alphanumeric-compatible system like netsend/netreceive that also 
supports broadcasting.

We implemented disis versions of netsend/receive that combine netsend and 
udpsend into one (thus allowing broadcasting udp packets while also allowing 
text and numbers). I'll be releasing code for these and other externals as soon 
as I get a chance to clean them up for distribution.

These however still don't cut it (or to be more precise are kludge to use) when 
having multiple clients. This is where in theory netclient/netserver should 
shine, except for the aforesaid issues.

I've thought of however a way of significantly lowering broadcast's overhead as 
currently we broadcast cues for all musicians to all clients, rather than 
parsing them out (which is something that netserver conveniently includes). So, 
this may hopefully also cut down on the amount of data being sent.

Come to think of it, perhaps having to broadcast the same thing to multiple 
clients still fills up the same buffer causing problems? Is this indeed so? 
(e.g. sending a same text to 27 different clients, does that effectively 
multiply its size 27 times within the context of 4096 limitations).

Finally, it appears that netserver and netclient have arbitrarily assigned 4096 
limitations with a DEFINE inside their source. I am wondering is this because 
of TCP's architecture (I thought I heard somewhere it could do up to 65535 per 
packet) or is this something we could simply change and recompile and thus 
achieve much better results?

Please advise.

Best wishes,

Ico


_______________________________________________
Pd-list@iem.at mailing list
UNSUBSCRIBE and account-management -> 
http://lists.puredata.info/listinfo/pd-list

Reply via email to