[colug-432] protocols (Re: SSH)

[Jeff Frontz]

Don't get me wrong-- I get the niceties of text-based protocols and
make use of them all the time.  I'm just the curmudgeon who has to
make things run faster with fewer memory/cpu/network/bus/disk
resources and without a servlet running in a virtual machine to make
everything pretty for me; determinism in protocols is my friend.

Jeff

P.S for the curious reader who likes to read the whinings of a bit twiddler:

Take a look at SIP over TCP (SIP is one of my oldest and bestest
text-based frenemies); it's what everyone is moving toward (if not
already using) for telephony (be it traditional phone calls or
"presence") and the big boys are using it over TCP (for reliability).
There is no away a priori to know how big of a buffer you're going to
need to slurp an INVITE.  You have to do byte-by-byte inspection to
know when you can stop (or you can do something nasty like this guy
suggests: https://lists.cs.columbia.edu/pipermail/sip-implementors/2006-July/013611.html
).  Even in UDP land (where you have some nominal MTU size to limit
things), you have no idea where in the SIP message the parts you are
looking for might be found (or if they're even going to be there).

Compare that to, say H.323 or SS7/MTP -- or TCP or IP or UDP -- you
know exactly how big the message will be and can pre-allocate buffers
accordingly (and you know exactly where to look for screening/routing,
etc.).

Similarly, I don't think anyone would want to use "web services" to do
their disk I/O, whether it's over the LAN or across the bus.  Heaven
help us if (when?) there would ever be an IETF RFC purporting to be a
"replacement" for USB! <shudder>



On Tue, Mar 8, 2011 at 6:07 PM, Richard Troth <rmt at casita.net> wrote:
> Sure, arbitrary length needs conscious handling.  HTTP did that already.  So
> did SMTP.  Many ways to do it.
>
> A really sweet one is XMPP.  It's one byte guzzling XML stream.  But it's
> readable (sort of).
>
> With plain text protocols, you can write your server as a shell then throw
> it under 'xinetd' or 'inetd' and dispense with the TCP handling.  Ahhh...
>
> This stuff is gorgeous.  I gotta make you a convert, Jeff.   :-)
>
> -- R;   <><
>
>
>
>
>
> On Tue, Mar 8, 2011 at 17:56, Jeff Frontz <jeff.frontz at gmail.com> wrote:
>>
>> Plain text protocols are (often) nice for debugging, but they're
>> painful to write robust software around (just how to you make sure you
>> can handle everything about a protocol that has a message that can be
>> any arbitrary length and that gives no clue as to how long it might
>> be?  SIP anyone?)
>>
>> Jeff
>>
>> On Tue, Mar 8, 2011 at 5:51 PM, Richard Troth <rmt at casita.net> wrote:
>> >  I honestly love plain text protocols.  Wish we had more of
>> > them, then cover with SSL when needed.
>>
>> _______________________________________________
>> colug-432 mailing list
>> colug-432 at colug.net
>> http://lists.colug.net/mailman/listinfo/colug-432
>
>