[mnet-devel] Re: [mnet-cvs]kb-scaled timeouts

[Zooko]

> > Oh -- firstmsgbody isn't mencoded yet?  Well, the way is
> > "bytes = len(mencode(firstmsgbody))".
> 
> feel free to change the code in any way you think makes sense.

I don't have time to make this change today (or more precisely: time to test it 
and commit it), as I'm moving into the new house.

It isn't really buggy the way it is, but len(pickle(msgbody)) is not *exactly* 
the same as len(mencode(msgbody)).

Hm.  In fact, the *real* number you want is len(wholemessage), where 
wholemessage is msgbody plus headers, mencoded and compressed with gzip.

No wait, in fact the *real* number you want is len(wholemessage) + 
len(wholeresponsemessage), because if I send a message like "please tell me your 
new IP address", versus I send a message like "please send me a 1 MB block", 
then the turn-around time doesn't correlate at all with the size of the 
*request* message alone!

> i would prefer if communications still worked after the changes,
> which it did not anymore after increasing the blocksize.

What happened?  We started using blocks > 256 MB, and v0.6.1 nodes had comm 
exceptions due to messages too large ?

--Z



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
mnet-devel mailing list
mnet-devel at lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mnet-devel