In a message dated 09-01-96 RICHARD TOWN wrote to JIM PALMER:
RT> Thanx for the info.  But I don't know of any transmission speed that will
RT> support 8: 1 compression on the fly.
Nor do I, V.42bis is not a speed but a data compression protocol.  It may be 
applied to any speed modem that runs V.42 error correction.  It only 
increases 
the speed of 'compressable' data.
RT> =  0:22:55  CONNECT 21600/ARQ/LAPM/V42BIS
RT> *  0:24:31  Rcvd C:\BBS\FD\INBOUND\1X30.TST; 983040b, 10685 CPS
As I read this, your modem got a full 4:1 compression on this file, but 
10,685 
cps is very close to the 115,200 bps of your UART.
RT> Are you really saying that this file (let's keep to the same standard)
RT> will achieve a faster apparent effective thruput just because your
RT> serial port/modem can work at 230k4?
Not if it is uncompressable, (already ZIPped) even a 38.4 kbps port can keep 
up.  If it were plain text, (even screens) yes.
RT> If so, then that would be news to Joe F. at Skokie who reports:
RT>         --------------------------------------------------------
RT>   Speed | 1X30.TST | 2X10.TST | 3X06.TST | 4X04.TST | 5X16.TST |
RT>  ---------------------------------------------------------------
RT>   14400 |   10077  |    3974  |    2339  |    1710  |    6107  |
RT>   16800 |   10867  |    4635  |    2724  |    1996  |    7117  |
RT>   19200 |   10959  |    5260  |    3096  |    2283  |    8062  |
RT>   21600 |   10972  |    5983  |    3502  |    2565  |    9125  |
RT>   24000 |   10980  |    6635  |    3899  |    2851  |   10121  |
RT>   26400 |   10987  |    7153  |    4191  |    3135  |   10318  |
RT>   28800 |   10980  |    7737  |    4571  |    3414  |   10443  |
RT>   31200*|   10994  |    8414  |    4964  |    3700  |   10654  |
RT>   33600*|   10954  |    8904  |    5279  |    3983  |   10701  |
RT>  ---------------------------------------------------------------
RT> Do you have any log extracts showing this file being transmitted any
RT> serially faster via POTS?
Gosh no, my Mac software just makes dialog boxes, sorry.  However your 
chart can show the effect of a slow UART.
Let's do the math:
If we have a stock V.34 modem, no V.42, we can theoretically acheive 2880 
s.
If we add V.42, we can get 3600 cps in theory, lower in practice, due to 
packetizing. (I suspect this is column '4X04.TST' above) Note that a 38.4 
kbps 
serial port will still not impede this.
Now, if we use V.42bis with 4 to 1 compression our theoretical maximum 
becomes 
14,400 cps, (your column 5X16.TST may represent actuals observed with 
V.42bis) 
but only if our DTE speed is at least 153,600 kbps.
As the '5X16.TST' column does not increase by the 16% that the others do from 
28.8 to 33.6, the test suite above must have used a UART at 115.2 Kbps to get 
those results, had they used 230.4 kbps devices, (both the modem and the 
computer) I would have predicted about 14000 at 33600 in column '5X16.TST'.
If you enter all the above data (except'1X30.TST') in a spreadsheet and graph 
using 'Speed' as the x-axis and you can see him over-run his UART above 24000 
in the column '5X16.TST' only. There is a big 'bend' in the data, which 
levels 
out just below the UART speed.
Finally, if our version of V.42bis allows 8:1 compression our max becomes 
23040 cps at 28.8, but only if we can open the DTE speed to 230.4 Kbps. (It 
would be 28,800 cps with a fast enough UART, I am 'limited' to 230.4 k)
(Picky disclaimers: when testing modems, one can only find the limits of the 
slowest device.  Getting a fast modem, UART, CPU, etc. will only help you 
identifiy the slowest component your sysop owns.  When exceeding 9600 bps, 
one 
also gets the wonderful chance to see how much thought your phone company has 
put into line quality.  ZIPped files are rarely further compressable.  RPI 
modems are suitable as doorstops.  Your smilage may vary.)
---
 * TIMM 1.0.6  * Whoever named it 'necking' had a poor grasp of anatomy.
--- WILDMAIL!/WC v4.12 
---------------