Path: utzoo!attcan!uunet!pcrat!rick
From: rick@pcrat.UUCP (Rick Richardson)
Newsgroups: comp.dcom.modems
Subject: Re: V.32 will dominate the marketplace (Was: Re: Which is best?)
Keywords: Packet Forwarding, X.25, ISDN
Message-ID: <617@pcrat.UUCP>
Date: 12 Nov 88 14:17:06 GMT
References: <407@telly.UUCP> <1245@nusdhub.UUCP> <422@scifi.UUCP>
Reply-To: rick@pcrat.UUCP (Rick Richardson)
Organization: PC Research, Inc., Tinton Falls, NJ
Lines: 58

In article <422@scifi.UUCP> njs@scifi.UUCP (Nicholas J. Simicich) writes:
>Interesting point. Now, there are two reasons that a modem like a Trailblazer
>is slower in "negotiated" full duplex mode.  One is the fact that if it is
>transmitting, it can't be receiving (hmmm...I think the same is true of my
>Unix application, unless I do complex things with multiple processes and IPC)
>while the other thing is that when you write to the modem it must wait a short
>period of time before it decides that no more characters are coming and closes
>the packet with a CRC.

There are some clever ways to minimize the packet forwarding delay, not
blessed by the CCITT, but in AT&T's ISDN terminal adapters.  You can get the
delay down to just a few milliseconds.

>
>Furthermore, there is an additional effect.  Modems running MNP at any speed
>must collect a full MNP packet, internally, and decide that they have it
>correctly before they can transmit any of it to the attached computer.  At least,
>this seems to be true through level 3.  You might tell me that MNP is optional
>for a V.32 modem.  I consider this to be a drawback, not a benefit.  Because

This delay dominates in protocols, such as X.25 used in ISDN, when the
(user, file transfer, transport layer, what have you) protocol is not
windowed (Such as XMODEM, YMODEM, old kermits).  End to end, you see
total delays of approximately:

	For an XMODEM data packet:
	PL * Tx RS-232 character time		132*.52ms	(19.2kbps)
	Tx Forwarding delay			5ms
	PL * Tx Comm. Medium character time	132*.125ms	(64kbps)
	PL * Rx RS-232 character time		132*.52ms	(19.2kbps)
	
	For the return acknowledgement:
	PL * Tx RS-232 character time		1*.52ms
	Tx Forwarding delay			5ms
	PL * Tx Comm. Medium character time	1*.125ms
	PL * Rx RS-232 character time		1*.52ms

	TOTAL:					165ms

Thats only about 6 packets/second or 768 characters/second.  To get
decent speed, you either have to change the file transfer protocol
(users resist this), or change the underlying protocol between the
terminal adapters (be they modems, or ISDN TE's).  In the V.32
world, you "send and pray".  In AT&T's ISDN world, you choose DMI
Mode 2, a rate adaption scheme that makes the 64k channels look
like pieces of wire.  Also "send and pray", but with very low claimed
bit error rates.  Of course, the file transfer protocol will recover
from errors, anyway (hopefully), so no great loss.

The point is, your "drawback" can be of significant economic benefit.

I *am* happy with the 'blazer.  The spoofing is one way to minimize
these problems.  But it isn't a panacea, nor are V.32 or ISDN.
-- 
Rick Richardson | JetRoff "di"-troff to LaserJet Postprocessor|uunet!pcrat!dry2
PC Research,Inc.| Mail: uunet!pcrat!jetroff; For anon uucp do:|for Dhrystone 2
uunet!pcrat!rick| uucp jetroff!~jetuucp/file_list ~nuucp/.    |submission forms.
jetroff Wk2200-0300,Sa,Su ACU {2400,PEP19200} 12013898963 "" \r ogin: jetuucp