VishwasM at netplane.com
Mon Jun 23 07:14:12 PDT 2003
I was reading thru the document and had a few points to make.
1. If we are talking about RFC2119. The terms should be capitalized(thats
the norm in the IETF).
"The terms "may," "must," and "shall" are used as described in RFC 2119,
2. I am not sure where the term 80 ms for signalling servers came from
shouldn't it be 60ms, in context of SONET.
3. RFC2313 has been obsoleted by 3513. Also draft
draft-ietf-ipv6-unicast-aggr-v2-03.txt changes further, though not sure of
implementation in this case.
4. Also if we have RFC2461, shouldn't we also try to do 3122 also. Also
5. For SNMP we may want to refer to documents in the
http://www.ietf.org/html.charters/snmpv3-charter.html page. Also RFC1157
6. Regarding "Hot Insert" should provisions be made for a seperate card
insertion from what was already there before. I mean replacing with a
non-similar card, though with some limitations.
7. In the serviceablity section in the resource monitoring section. I feel
we could have requirements of: -
a. Actually allowing a per thread CPU limit as provided in WSRM.
b. Allowing a configureable per subsystem memory utilization, not just one
on a system basis. Most new equipment vendors seem to do this.
c. Also dampning to prevent frequent thrashing should be provided too, like
having low and highwater marks.
8. Tools for debugging. I see requirements for tools for showing all valid
variables in a current context, in windows its called "variables".
9. I think ability to provide stricts priority scheduling is also a
requirement. Also a need to provide a large number of priorities to threads.
More information about the cgl_discussion