home *** CD-ROM | disk | FTP | other *** search
- Path: sobt.accessorl.net!user
- From: eric@accessorl.net (Eric Shaw)
- Newsgroups: comp.dcom.modems
- Subject: Re: To V42Bis or not to V42Bis??
- Date: Sun, 04 Feb 1996 20:03:49 -0500
- Organization: Access Orlando
- Message-ID: <eric-0402962003490001@sobt.accessorl.net>
- References: <4ej6rh$oa2@canopus.cc.umanitoba.ca> <4eruq2$e2@hg.oro.net> <4etjse$a0g@azure.acsu.buffalo.edu> <4f0jnq$dr7@hg.oro.net>
- NNTP-Posting-Host: sobt.accessorl.net
-
- In article <4f0jnq$dr7@hg.oro.net>, estarry@oro.net (Ed Starry) wrote:
-
- >~the extra overhead involved in compressing files can slow down
- >~your throughput.
- >
- > If you really believe this you had better get your calculator
- fixed. When I
- >see 7,000 CPS throughput rates I don't care how much 'extra overhead' there
- >is, 7,000 is 7,000! I always thought 7,000 CPS was faster than 2,880 CPS?
-
- In the case of compressible files that can be sent at 7000 cps, you are
- obviously right. That is not what I was refering to. If a file is only
- slightly compressible, i.e. the best compression engine would give you
- 3500 CPS, then a slower compression routine could take extra overhead,
- bringing you below the rate you'd get with compression off. For example,
- you might get 3388 CPS with compression off, but only 3300 with
- compression on if the file is not very compressible, not because v.42bis
- is making the file bigger, but because the modem took too long to figure
- out that it couldn't compress it well. This doesn't happen often, and
- the slowdown is usually so small that the parts of the file that compress
- a little quicker counter it, but it is possible for it to be a tiny bit
- slower with compression. (with MNP5 it is A LOT slower with "compression",
- but thats for other reasons)
-
- Note: the 3388cps without compression comes from dividing 28800 by 8.5
- bits/byte instead of 10. Error correction strips out the start and stop
- bits between the modems, so there are 8 bits per byte instead of 10 like
- you have through the serial port. I used 8.5 instead of 8 then because
- there is a little bit of overhead for the v.42 checksums. The number is
- not *exactly* 8.5, but it is close. The exact number depends on your v.42
- packet size.
-