michel@BBNCCT.ARPA (Tony Michel) (01/13/86)
I routinely use Atari Kermit with a UNIX-like system (a BBN C/70) via a TAC (version 113). First - set the intercept character to be something Kermint doesn't use. I use (ctl-F). Do this with the sequence "@i 6." This sets the TAC commmand interpreter recall character to something you can still type, and therefore still get back to it. If you say "intercept none (@i n) you can't get back. Next - enable flow control with two commands (@ f i s and then @f o s) which start input and output flow control. These two sequences enable XON/XOFF flow control, which allows the TAC to temporarily shut down the Atari to TAC direction, which is usually the bottleneck. The usual problem solved by flow control is that the TAC terminal input buffer is usually set to 64 characters. Since the default KERMIT buffer is about 90 characters, it is easy to overrun the TAC. A second strategy here is to set the KERMIT block size to something less than the TAC input buffer size (less than 64). I have heard rumors of some UNIX KERMIT servers which negotiate you back up above 64 characters, unbeknownst to the user, so try the flow control dodge if small blocks don't work. There is also a penalty to small blocks - lower throughput. Unless youknow you are doing binary data transfer, which I haven't tried with my KERMIT (I don't think it does it), don't fool with the TAC commands @b i s and @b o s. I don't think they do anything helpful for you. I use the standalone version of Atari 800 KERMIT written in ACTION! by Jack Palevich. It seems flawless. It also has a nice debug mode in which it prints out everything that's going on. The sources floated by on info-atari many moons ago. It seems to need the cartridge, which is fine with me. I use the above aggregation on a DOS-XL system, with a an R-Verter RS-232 port instead of an 850 box. AXM P.S. Note that in the above, once you have re-defined the intercept character, you must type the new intercept character where you previously would have used "@."