home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.dcom.modems:11267 can.uucp:181
- Path: sparky!uunet!wupost!cs.utexas.edu!torn!cunews!revcan!ecicrl!eci386!woods
- From: woods@eci386.uucp (Greg A. Woods)
- Newsgroups: comp.dcom.modems,can.uucp
- Subject: Re: UUCP 'g' vs. MNP & V.42
- Message-ID: <1992Jul27.193824.3733@eci386.uucp>
- Date: 27 Jul 92 19:38:24 GMT
- References: <1992Jul23.180658.20724@eci386.uucp> <mN8FoB2w164w@zswamp.UUCP>
- Organization: Elegant Communications Inc.
- Lines: 54
-
- In article <mN8FoB2w164w@zswamp.UUCP> geoff@zswamp.UUCP (Geoffrey Welsh) writes:
- > Just the facts so's the folks make the best decisions:
-
- Remember that unless you understand the relationships between
- different protocols being used (i.e. how they interact), and the
- effects that may be due to particular implementations or due to
- parameters set in particular connections, you cannot use "common
- sense" in making statements about communications issues, even if you
- know many of the details about each protocol.
-
- > woods@eci386.uucp (Greg A. Woods) writes:
- > > MNP-4 is link-layer protocol that retries erroneous packets. UUCP 'g'
- > > is a link-layer protocol that retries erroneous packets. Guess what
- > > -- the don't get along very well!
- >
- > This is *completely* untrue! While MNP may increase latency compared to a
- > normal link, it does not in any way interfere with the data being sent.
-
- Except where that latency will cause problems with some other file
- transfer or error correcting protocol that's being encapsulated by MNP-4.
-
- We're talking about UUCP 'g' here, and with "normal" parameters,
- uucico will often wait (and sometimes timeout while waiting) for
- ACKs. This degrades throughput. It is effectively a loss of ~15%
- when MNP-4 and UUCP 'g' are used over V.22bis, and as much as 50% when
- used over V.32. These are measured results.
-
- > There is no harm in using an error-correcing protocol over an 'error-free'
- > link (I put that in quotes because MNP guarantees no errors between the
- > modems, not no errors between the computers), unless the protocol imposes a
- > significant overhead. Since, in this case, the overhead is more than made up
- > for by start and stop bit removal, there is no reason not to use MNP under
- > UUCP-g, at least not at 2400 bps where the increased latency should not push
- > past the window size.
-
- Sorry, but that is *NOT* the case, as measured in actual connections
- with UUCP 'g' window size at both 3 and 7. Even the 20% gain in
- efficiency courtesy of MNP-4 is still not sufficient to overcome
- timing problems.
-
- [No, at 2400 bps the increased latency of MNP-4 without spoofing does
- not usually cause *timeouts* with UUCP 'g' at a window size of 3,
- packet size of 64, but will cause delays. Note however that line
- noise will immediately increase the latency, and thus the chance of a
- timeout. So, spoofing is a double win -- 1) the host-to-modem link is
- guaranteed error free; 2) the DCE-DCE link latency does not affect the
- file transfer protocol.]
-
- So, any other questions about why UUCP sites everywhere use Telebits? :-)
- --
- Greg A. Woods
- woods@Elegant.COM, woods@robohack.UUCP VE3TCP ECI & UniForum Canada
- +1-416-443-1734 [h] +1-416-595-5425 [w] Toronto, Ontario; CANADA
- Political speech and writing are largely the defense of the indefensible-ORWELL
-