On 6/30/06, Ed Tomlinson <edt at aei.ca> wrote:
>
> I might be tempted to agree with you if 0.7 was not alpha.  The bugs found
> can easily be serious enough to really affect the the workability of freenet.
> We now have auto update.  So timed mandatory builds make a lot of sense.

So which recent mandatory builds have been fixing problems so bad that
only accepting 1/10 as many requests from old nodes would still let
those old nodes damage the rest of the network?

IIRC (and I only read the brief blurbs about what the mandatory builds
fix, not anything detailed), all the recent things have been about
load balancing, backoff, etc -- basically some manifestation of either
the old build sends too many requests, or the old build is somehow
really bad at answering requests.  In either case, what's the harm in
accepting a very occasional request from it?

And, in the extreme case where the above does not apply, and accepting
a couple requests would be bad (I don't see how it could, but I'll
grant it might happen), then just do a timed self mandatory build like
the current ones.

I guess what I'm saying is, for the majority of mandatory builds, I
believe my proposal would help a few people a lot, a fair number of
people a little (convenience of not having to run update.sh just
because you left town for the weekend and turned off your computer),
and hurt no one.  The only reason not to implement it that I can think
of (assuming the option remains of a truly mandatory build) is if it
would take a lot of time to code.  But that would surprise me.  (Of
course, I haven't written a line of Freenet code, and I've read only a
tiny bit of it, so take my assumptions about that with an appropriate
grain of salt.)

Just because it's alpha doesn't mean we shouldn't try to be nice to
users when it's easy to do so.  It just means that if we *have* to
break things, or think it would be good to for some reason, that the
users shoudn't complain about not being warned.

Evan Daniel

Reply via email to