[tahoe-dev] Multiple introducers

[Chris Goffinet]

Great! Thanks!

I haven't dived into the source yet but was curious, as you start  
having > 100 connections, I know your using Twisted framework, any  
thoughts on using epoll reactor if its not used when you start having  
higher load?

Was just curious on that.


On Apr 14, 2008, at 9:15 PM, zooko wrote:
> On Apr 14, 2008, at 9:53 PM, Chris Goffinet wrote:
>> I just dived into Tahoe and so far it seems really neat.
>
> Welcome!
>
>> I had one
>> question. What happens when the system that is running the introducer
>> goes down? The peers can't connect. How could one go about setting up
>> multiple introducers and having the clients connect to any random
>> introducer that would be up without editing the introducer.furl ?
>
> This is a good question.  We've cooked up a scheme to do this
> ("decentralized introducers"), but we haven't gotten around to
> implementing it, since a single introducer is good enough for now for
> the allmydata.com backup service.
>
> Here's the ticket:
>
> http://allmydata.org/trac/tahoe/ticket/68#comment:11
>
> (Skip to the end of the ticket to see the simple scheme that I
> currently want.  There are a few other ideas along the way that are
> no longer uppermost on my mind.)
>
> If some good Python hacker out there wants to start adding
> improvements to Tahoe, I would be very happy to see them start with
> this ticket.  Be aware that we don't accept patches into trunk
> without unit tests and documentation.  If you do get a patch into
> Tahoe then you get your name in the CREDITS file.
>
> Regards,
>
> Zooko
>
> _______________________________________________
> tahoe-dev mailing list
> tahoe-dev at allmydata.org
> http://allmydata.org/cgi-bin/mailman/listinfo/tahoe-dev