[mnet-devel] What's left to be done for 0.7

[Jim McCoy]

On Friday, October 3, 2003, at 10:49 AM, icepick wrote:
>  * Should we stop and convert over to Twisted?
>

Wasn't this how the current set of 0.7 screw-ups got started?  With a 
"should we stop and replace the MT with some unknown, untested research 
project" sort of query that started to snowball....

I think that eventual migration to twisted is a good thing, but if mnet 
ever does another "let's stop and break everything while we conduct 
some research" again it will descend so far into irrelevance that by 
the time you are finished no one will care. What is needed is a plan. 
Twisted offers some good benefits, but it would be better to figure out 
where it is easiest to start applying it and work from there instead of 
attempting some sort of massive conversion that is bound to fail.  
(Why? Because eventually some key member of this transition team will 
have some "real life" event stop them from making progress and things 
will block on that person's area, and then it will happen to another 
person, etc.  This is what happened with ent, if people will 
remember....)

Zooko made a good suggestion a while back that the easiest transition 
to twisted would be to start with hollowing-out the DoQ and replacing 
its internals with a twisted reactor.  This doesn't break existing 
things which expect a DoQ interface but lets refactored code talk to 
the reactor directly.  This is also a first step towards converting 
EGTP into a twisted protocol (which is a better solution than 
twisted.pb) and taking advantage of other twisted resources.

Jim



-------------------------------------------------------
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