[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Rules Language


Folks,

the email below seems to be the last email/action we saw on the rules language. This was more than two months ago. Given that we're already far behind our deliverables, it seems logical that we suggest to remove this work item from the charter.

Unless someone volunteers by Friday, 10/28/2005, 5pm EDT to take ownership of this work item and make progress before Vancouver, I'll ask our AD and the IESG to remove this work item from our charter. [Please note, that we already delayed this work item from the first charter].

Thanks,
  Markus

Alex Rousskov wrote: 
On Tue, 2005-08-09 at 22:58 -0600, The Purple Streak, Hilarie Orman
wrote:

http://www.purplestreak.com/ietf-opes/draft-ietf-opes-rules-language-hopalong-00.txt

1. Applicable to partial messages (packet-by-packet) 

I would remove the references to "packets" from the draft. None of the
protocols we think of dealing with (HTTP, OCP, SMTP, IMAP, RTSP) have a
notion of a packet.

Support for partial messages or, more precisely, partial message
payloads (potentially large bodies and such) is a must, of course.
Beyond that, incomplete "parts" should be invisible to the rules
language. The language does not need to deal with a truncated HTTP
header field or a partially received IPv6 address, for example. 

If the protocol defines an "element" or "part", the rules should be able
to access that part (usually via protocol-specific modules). Splitting
e;ementary parts into smaller chunks should be out of scope.

Note that if your description promises support for application protocol
elements (headers, bodies, etc.) than it would be possible to support
packets as well, but only for protocols that deal with packets...


Another unrelated comment is that we should not assume that a message is
always headers+body. Even HTTP messages are not like that because they
may have trailers. I think a couple of places in the draft may imply
otherwise.


If there is an error but there are no matching onError rules, what
happens?

Is there arithmetic support? For example, can I count the number of
certain headers in the message?


Please add at least one non-trivial/semi-complete realistic example.

Thank you,

Alex.