home *** CD-ROM | disk | FTP | other *** search
- Path: mips.pfalz.de!not-for-mail
- From: naddy@mips.pfalz.de (Christian Weisgerber)
- Newsgroups: comp.dcom.modems
- Subject: Re: Question about Error Correction and Compression
- Date: 13 Jan 1996 01:16:20 +0100
- Message-ID: <4d6tkk$npr@mips.pfalz.de>
- References: <4csame$orf@hermes.acs.unt.edu>
- NNTP-Posting-Host: mips.pfalz.de
-
- bryon@jove.acs.unt.edu (Bryon Sutherland) writes:
-
- > Many different applications are timing dependant and need error correction
- > and compression turned off,
-
- "Many" is arguable. Anyway, are you sure *your* applications are among
- those?
-
- > but it seems logical to me that only one modem (on the server end
- > preferably) needs to have them turned off.
-
- Basically right.
-
- > Won't the modem that's dialing in talk-down to the serving modem and not
- > use error correction and compression?
-
- If not configured to behave otherwise, yes. A potential problem is that
- error control is negotiated in-band, and the negotiation is initiated by
- the caller. Assuming that the modem is a modern one with V.42, and
- configured for "auto-reliable mode" (a fair assumption nowadays) the
- following will happen:
-
- 1. The calling modem will try to establish a V.42/LAPM connection. It
- will send a stream of XONs of alternating parity (i.e. 0x11 0x91 0x11
- 0x91 ...) and wait for an acknowledging "EC".
-
- 2. If it doesn't receive a response, the modem will try to establish a
- connection according to the V.42 "Alternative Procedure", better
- known as MNP. For this is will send an MNP LR frame, i.e. a bunch of
- binary garbage characters.
-
- 3. If it doesn't receive an MNP LA frame in response, the modem will
- give up and run the connection without error control.
-
- Note that different brands of modems show different degrees of
- cleverness and persistence for this negotiation, both at the calling and
- answering end. Steps (1) and (2), respectively, are usually repeated a
- few times each before falling back to (2) and (3), respectively. Modems
- with MNP10 will be more aggressive in their attempts to establish an
- error control link.
-
- I *think*, answering USR Couriers with error control disabled (&M0) will
- try to abort incoming requests for error control (answering "E\0" at
- step (1) etc.) as quickly as possible.
-
- > If I could set my servers modems that way I wouldn't have to worry about
- > each of 300 users properly configuring their modems, right?
-
- From the description above, it should be clear that at the answering
- end, with error control off, you'll receive a stream of garbage
- characters and it will take the calling modem a few seconds to fall
- back. Data sent during this time to the caller may be lost. Whatever
- your applications are, they'll have to deal with all this.
-
- Many years ago, when I was new to modems, networking, etc., I
- participated in a small German BBS network (MagicNET, for those German
- readers who remember). The BBS software we used prompted the caller to
- press return, ostensibly to determinate the parity type used, before
- continuing with the login. If, however, a special character, I think it
- was ^F, was entered instead of a carriage return, the BBS would drop
- into netcall mode, which would't allow a user login anymore. (Yes, all
- of the design was utterly braindead. It wasn't mine, but I'll admit that
- I wouldn't have known better back then, for that matter.)
-
- When modems featuring error control became popular, it turned out that
- their MNP negotiation attempts always caused the BBS to drop into
- netcall mode, if it didn't have an MNP-capable modem itself. This caused
- a lot of confusion until people realized they had to disable error
- control. Due to the limitation of the BBS software and netcall protocol,
- their was nothing that could be done at the answering end.
-
- --
- Christian 'naddy' Weisgerber naddy@mips.pfalz.de
- See another pointless homepage at <URL:http://home.pages.de/~naddy/>.
- -- currently reading: John Norman, Dancer of Gor (#22) --
-