Balazs Scheidler wrote:
> My reasoning is:
> * solve possible TCP interop problems with current implementations
> * to provide a way to send reliable messages for small edge devices which
> probably will never implement SCTP nor BEEP (DSL routers for instance)
Well, during the early days, I tried to push for a better timestamp and
it was considered a modification to the payload. Of course, the cooked
mode does a better job of carrying the timestamp, but that's admittedly
a lot more overhead. However, I'm not the authoratative answer on that.
:-)
I don't think you'll get something standardized from scratch and
deployed faster than you can get a BEEP implementation deployed. Plus,
for 99% of the uses of BEEP, you're only using one channel at a time
anyway, so whether it's TCP or SCTP is kind of irrelevant.
As for DSL routers, the ones I've seen have a web server for
configuration and status and such, with plenty of oomph for running a
simple (not very configurable) version of syslog-reliable, BEEP and all.
Once the initial connection is established, you're really doing almost
nothing to actually send the messages, at the same level of effort as
tracking a TCP window.
--
Darren New
San Diego, CA, USA (PST)
"I'm allergic to antihistamines."
"Oh? What do you break out in?"
