
On Wed, Oct 24, 2007 at 07:32:05AM -0400, Gwern Branwen wrote:
On 2007.10.24 05:50:39 -0500, Spencer Janssen
scribbled 0 lines: On Tuesday 23 October 2007 19:40:04 gwern0@gmail.com wrote:
Tue Oct 23 20:13:41 EDT 2007 gwern0@gmail.com * Run.hs, SshPrompt.hs, ShellPrompt.hs: mv runInXTerm back into Run.hs per suggestions
Tue Oct 23 20:16:28 EDT 2007 gwern0@gmail.com * Run.hs: +my suggested runInTerm general function
Tue Oct 23 20:18:56 EDT 2007 gwern0@gmail.com * Run.hs: specialize runInXTerm to use runInTerm per my mailing list suggestion
Tue Oct 23 20:39:11 EDT 2007 gwern0@gmail.com * Run.hs: do my usual segregation into safe and unsafe runInTerms
Applied.
I don't like these "safe" and "unsafe" names, please consider something more descriptive.
Cheers, Spencer Janssen
Does anyone have any suggestions? I originally chose safe/unsafe because I regard not going through the shell as safer, less error-prone. What would be better? 'shell'/'shelless'?
I'd definitely use the word "shell" in the name. Using a shell really is no less safe than using no shell, in my opinion, unless you are using potentially hostile input. But if we're exec-ing potentially hostile input, it takes a lot more than passing the arguments in verabatim to ensure that the resulting action is safe. How about something like: execShell and exec, which describe what is actually being done? Or just append ViaShell to the functions that go through the shell? -- David Roundy Department of Physics Oregon State University