[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Requirements
I have thoroughly read through the pilc working group material. It is mostly
survey material with discussions on TCP's reaction to layer 2 activity, TCP
versions, and potential changes to TCP and options. The focus is on TCP in the
face of lower layer behavior. I don't recall seeing any discussions from the
other perspective of what kind of behavior should layer 2 be limited to ...or
what makes a good layer 2 for IP? This in my mind is the interesting wireless
(RAN) question. Regards John
"Dana L. Blair" wrote:
> Snip from James' response.
> > One useful thing would be to try to get a handle on what makes
> > a "good" L2 for IP. We have tried to do that for handoff in
> > draft-manyfolks-mobilereq-ipv6-00.txt, which includes requirements
> > for an L2 trigger API or protocol.
>
> The Performance Implications
> of Link Characterists (pilc) working
> group has spent alot of time describing the impact of various
> layer 2 techniques on TCP/IP.
>
> The pilc web page,
> http://www.ietf.org/html.charters/pilc-charter.html,
> contains some interesting drafts. There is even
> one about TCP over 2.5G and 3G wireless networks. I have
> not read all of these, but I know the slow links and
> links with errors drafts has some good info for link
> layer designers.
>
> End-to-end Performance Implications of Slow Links (41881 bytes)
> End-to-end Performance Implications of Links with Errors (38504 bytes)
> Advice for Internet Subnetwork Designers (92940 bytes)
> TCP Performance Implications of Network Asymmetry (103539 bytes)
> Link ARQ issues for IP traffic (47793 bytes)
> TCP over 2.5G and 3G Wireless Networks (33785 bytes)
>
> Request For Comments:
>
> Performance Enhancing Proxies Intended to Mitigate Link-Related Degradations
> (RFC 3135) (114825 bytes)
>
> thanks,
> Dana
>
> > -----Original Message-----
> > From: James Kempf [mailto:James.Kempf@Sun.COM]
> > Sent: Friday, July 20, 2001 7:28 PM
> > To: RRobin01@sprintspectrum.com; jgw@cisco.com; dblair@cisco.com
> > Cc: more@ops.ietf.org
> > Subject: RE: Requirements
> >
> >
> > Dana,
> >
> > While I agree with you in principle, I think a case can be made that
> > some wireless L2 protocols are better engineered for IP than others,
> > handoff not included.
> >
> > In particular, I think the 3G wireless protocols like wCDMA, which were
> > not originally designed for IP (but rather primarily for ATM-like
> > voice) are not as optimally designed as, say, 802.11.
> >
> > One useful thing would be to try to get a handle on what makes
> > a "good" L2 for IP. We have tried to do that for handoff in
> > draft-manyfolks-mobilereq-ipv6-00.txt, which includes requirements
> > for an L2 trigger API or protocol.
> >
> > jak
> >
> > >From: "Dana L. Blair" <dblair@cisco.com>
> > >To: "Robinson, Richard" <RRobin01@sprintspectrum.com>, "John
> > G. Waclawsky"
> > <jgw@cisco.com>
> > >Cc: <more@ops.ietf.org>
> > >Subject: RE: Requirements
> > >Date: Fri, 20 Jul 2001 17:42:59 -0400
> > >X-Priority: 3 (Normal)
> > >X-MSMail-Priority: Normal
> > >X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2615.200
> > >Importance: Normal
> > >
> > >My comments below.
> > >
> > >...
> > >
> > >> > it is just that the mobility of wireless subscribers as
> > well as some
> > >> unique
> > >> > performance requirements/problems in the air interface may
> > >> indeed bring
> > >> out
> > >> > special requirements - perhaps more so in the macro cell
> > >> environment of
> > >> > wireless operators than in controlled micro/pico cell links
> > >
> > >For the past 2.5 years I have been asking the question is
> > >there anything special about wireless that impacts application,
> > >transport, or IP protocols that is different than wired.
> > >
> > >Up til now, the only exception for wireless than wired that
> > >I have discovered is handoff. I believe that we SHOULD
> > >view the wireless access network as just another Layer 2.
> > >
> > >Here are some examples of suggested differences.
> > >
> > >1. Wireless links are special because they have low bandwidth
> > and large
> > >latency.
> > >
> > > Some wired links are slow and have large latency too.
> > >
> > > V.22 modems run over the PSTN and are slow with large latency.
> > TCP was
> > >designed to optimize bandwidth on slow and fast links.
> > > V.42bis, MNP compression, and Header compression were
> > invented several
> > >years ago to improve performance n
> > > wired links.
> > >
> > >2. Wireless links are special because they are error prone.
> > >
> > > Some wired links are error prone too.
> > >
> > > Many PSTN connections are error prone depending on wiring.
> > > MNP and V.42 error correction protocols were created
> > several years
> > ago to
> > >fix errors in the link layer.
> > >
> > > The RLP layer of cellular standards provides the same
> > > function.
> > >
> > >3. Wireless devices are special because they move.
> > >
> > > Roaming:
> > > Wired devices roam too.
> > > My laptop roams because it plugs into an ethernet at
> > > home or in the office depending on where I want to
> > > work.
> > >
> > > Handoff:
> > > One possible exception for wireless is handoff. There is
> > > really no need that I can think of for handoff in
> > > the wired world, but the Mobile IP WG seems to getting
> > > the right idea of Internet Handoff for wireless devices.
> > >
> > >4. Wireless devices are special because they need to know
> > where they are
> > -
> > >Location services.
> > >
> > > Wired elements devices can benefit from location services
> > > too, and be supported with a small inexpensive GPS chip.
> > >
> > >thanks,
> > >Dana
> > >
> > >> >
> > >> > some of these come through in Paul Reynolds draft - as
> > >> wireless operators
> > >> > get engaged in these requirements some of the things that
> > >> make wireless
> > >> > different may be highlighted - i'm reviewing the draft now
> > >> >
> > >> > Richard Robinson
> > >> > Sprint PCS
> > >> > 15405 College Boulevard
> > >> > Lenexa, Kansas 66219
> > >> > 913.890.4242 (fax 4100)
> > >> > MS - KSLNXZ0201
> > >> > rrobin01@sprintspectrum.com
> > >> >
> > >> > -----Original Message-----
> > >> > From: tim clifford [mailto:tjc@lacunanet.net]
> > >> > Sent: Monday, July 16, 2001 11:23 AM
> > >> > To: James Kempf; more@ops.ietf.org; jgw@cisco.com
> > >> > Subject: RE: Requirements
> > >> >
> > >> > which would seem to imply that we need to try to convince
> > >> people at the
> > >> ietf
> > >> > (i think its a misnomer to say "convince the ietf") that
> > we're talking
> > >> about
> > >> > more than a link layer, maybe the right term is
> > mobility, or roaming
> > >> > services, or disadvantaged user devices, or large populations
> > >> of always on
> > >> > subscribers ;-)
> > >> >
> > >> > tc
> > >> >
> > >> > > -----Original Message-----
> > >> > > From: owner-more@ops.ietf.org
> > >> [mailto:owner-more@ops.ietf.org]On Behalf
> > >> > > Of James Kempf
> > >> > > Sent: Monday, July 16, 2001 11:47 AM
> > >> > > To: more@ops.ietf.org; jgw@cisco.com
> > >> > > Subject: Re: Requirements
> > >> > >
> > >> > >
> > >> > > John,
> > >> > >
> > >> > > There is no such catalog. In general, IETF has been
> > resistent to
> > >> > > making wireless a special category. It is viewed as just
> > >> another link
> > >> > > layer.
> > >> > >
> > >> > > jak
> > >> > >
> > >> > > >Date: Sat, 14 Jul 2001 11:15:34 -0400
> > >> > > >From: "John G. Waclawsky" <jgw@cisco.com>
> > >> > > >To: more@ops.ietf.org
> > >> > > >Subject: Re: Requirements
> > >> > > >
> > >> > > >Does anyone know if the IETF does an cataloguing of activities
> > >> > > with regards to
> > >> > > wireless?
> > >> > > >For example is there a cross reference anywhere that
> > >> describes wireless
> > >> > > activities going
> > >> > > >on in the IETF? This information would probably be
> > very uesful
> > >> > > for the mobile
> > >> > > >operators. Regards John
> > >> > > >
> > >> > > >
> > >> > >
> > >>
> > >
> > >
> >