From: Daniel P. <da...@ri...> - 2002-01-26 03:02:56
|
On Fri, 25 Jan 2002, Kai Gro=DFjohann wrote: > Steve Youngs <yo...@xe...> writes: >=20 >> What is Tramp supposed to do if it can't find 'mimencode'? >=20 > I told it to barf. Not sure if that's the right thing to do. If the > remote end has Perl, then you can use the Perl implementation for > base64 encoding that comes with Tramp. This is done by using the > command "tramp_mimencode" for the method definition (in > tramp-methods). Likewise for tramp_mimedecode. >=20 > It might be an idea to have Tramp figure out whether it should use > mimencode, mmencode, tramp_mimencode, uuencode, or something else. > But that would make connection setup even slower, which people might > not be happy with. Thoughts? From testing, even over a slow link I found it possible to automatically determine a reasonable range of tools that were present on the remote end. The biggest killer for a TRAMP session is *latency* -- it hurts a lot less, even over < 1k per second links, to send a 2k shell-code executable as a lump than to wait for two or three round trip times. So, if you do want to do auto-detection, I suggest writing a posix sh compliant script that does the investigation work, editing it on the local machine before sending, then dumping it as a bulk lot to the remote machine and waiting for it's easily parsed response. ;) That way you get nicely tested auto-detection, progress -- and it's not tied to the performance of your network. It floods across, sits there and then reports what it's doing and what it finds. Regards, Daniel --=20 I am for an art of teddy bears and guns and decapitated rabbits, exploded umbrellas, raped beds, chairs with their brown bones broken, burning trees, firecracker ends, chicken bones, pigeon bones and boxes with men sleeping in them. -- Claes Oldenburg |