Designing socket messaging format
Manoj Plakal
terabaap at yumpee.org
Tue Nov 13 01:31:50 EST 2001
Steve Holden wrote:
> "Manoj Plakal" <terabaap at yumpee.org> wrote in ...
>> This might cover a lot of cases where you don't
>> need RPC semantics (no return value expected)
>>
>
> But some sort of response will usually be required, even if it's a simple
> ACK.
Sure, there will be situations where you
want an ACK, in which case you want
to do RPC. Or maybe you'd want an asynchronous
response instead of waiting for it. But
I believe there will be some cases where
you just want to send a notification and
not wait for anything.
>> and you're sending simple data values (so no
>> need of full-blown XML or other elaborate encoding).
>>
>
> How simple is simple? ASCII strings of less than 256 characters you might
> get away with...\
I was thinking about fixing a grammar and
then having a parser for this grammar rather
than having a general XML parser and giving
it the DTD for the protocol.
>> And you may not even need a reliable transport
>> protocol (so no need of HTTP or even TCP).
>> XML-RPC is cool but it seems like a really
>> big and overweight hammer.
>>
> Oh, so you don't mind if Word prints file "xyy.doc" when you ask it to print
> "xyz.doc"?
By reliable, I meant stuff like acknowledgments
and sequencing and all that. I assume that
the data payload itself is protected from
corruption. Doesn't UDP have checksums?
> Can we say "UDP?
Sure you can use UDP, but I don't want
to use a separate application-level protocol
on top of UDP for each app I write or talk to.
I'm not trying to reinvent a new network
protocol here. All I'm asking is if there's
some kind of cross-platform ultra-lightweight
COM/CORBA-kind of messaging system, built on top
of some IPC mechanism (UDP or domain sockets or named
pipes or something not too heavy).
I did find a link to XPLC: xplc.sourceforge.net
Manoj
More information about the Python-list
mailing list