Thanks Jean-Pierre,
I had a chance to make some of the changes before submitting to the IETF
(we had to go back and forth over minor issues.) We'll respond on the
list... Thanks,
raju
----- Original Message -----
From: GARBISU Jean-Pierre FTRD/DMI/CAE
<jeanpierre.garbisu@rd.francetelecom.fr>
To: <rajan@research.att.com>; <elizacelenti@att.com>; <sdutta@att.com>
Cc: <sls@ist-tequila.org>
Sent: Thursday, November 09, 2000 10:12 AM
Subject: [tequila/sls] comments on draft-sls-somefolks-00.txt
Hi
herafter you will find some comments and questions on your (very
interesting) draft-sls-somefolks-00.txt. It seems your STS/SIS structure
proposal could help a lot to rigorously define SLS.
best regards.
* 1 ( VoIP gateway example) : I would like to support the simple SIS
proposed. It seems very probable that the customer can't specify e.g.
transport delay or loss for such a service.
* 1 (3rd principle) : typo : Inasmuch.
* 3.C (1st §) : SLA (here and in several other places (e.g. in 3.D) is
mentionned instead of SLS, which doesn't seem necessary.
* 3.C (1st §) : typo : ascribes.
* 3.C.b : (in the SLS) instead of (in the SIS) ?
* 4 : could SLO definition be briefly reminded in the draft?
* 4.B1 (Number of SAP Items) : typo : the the.
* 4.B2 (the tag type / basic topological structures according to the
number of sources or destination) : could this relationship be presented
before in 3.B to help for the comprehension.
* 4.C1 : issue : will the rule "A larger Graph Identifier overrides a
smaller one" be sufficient to handle separetely QoS Units. Example if you
have a global ("mother") hose QoS Unit and various independant ("sons")
pipes QoS units, the various pipes Graph Identifiers won't have any
relationship between them. Perhaps a n-level numbering scheme would handle
it better?
* 4.C3 (DescriptorType) : why choosing Leaky bucket? Token bucket
seems more appropriate (for AF based SLS at least). Perhaps both could be
quoted. It seems the DiffServ PIB doesn't impose one or another.
* 4.C4 (Percentile) : you mention mean, max and min delay but only
define maxdelay before. Is min delay necessary?
* 4.C.4 (Jitter Descriptor) : millisecond scale could be indicated (as
for delay).
* 4.C.4 (Service Availability Descriptor) : (small) issue : are link
failure and node failure tolerance relevant in this SLS draft (which hides
the Provider network to its customer)? Is it a common practice?
* 4.D.2 (Delay Frequency) : is it necessary to fix the 100 ms units?
* 5.2 : proposal : to avoid ambiguity, could all parameters indicate
PS and CS on your example?
* 5.C.1 : I don't catch the role of Graph Identifier... Could you give
a brief example using it?
* 5.C3 : I don't catch the respective roles of MeanRate and PeakRate
here for the VLL example. Does PeakRate refer to available VLL bandwidth and
MeanRate to the PeakRate of the VLL SIS asked here?
jeanpierre.garbisu@francetelecom.fr
This archive was generated by hypermail 2b29 : Thu Nov 09 2000 - 20:59:27 CET