Eric Roman mailed out an interesting project that he heard about recently: rexec. Seems to be a new project under the old name for transparant and secure remote execution from the CS folks at Berkeley. Printed out the paper; it should make a good read.
More wedding things today; finalizing contracts with the Marriott (pizza-n-beer, yum), finalized numbers to Tippecanoe (rehearsal dinner, yum), table layout for the reception, etc. Getting down to the finer details -- T-9 days.
Spent the afternoon cleaning up the minime code -- I made bunches of changes to the socket and console routines to be able to write the shell client for the journal system (jjc).
Hmm.. just found an annoying bug in jjc: C-h C-h (i.e., hitting backspace twice) brings up the emacs -nw appropros list, but hitting C-g to abort the appropos list somehow makes jjc think that the emacs child has finished, and therefore jumps back to the prompt, but then seg faults and dies. Ugh! Gonna have to fix that one. :-)
Some guy mailed me today about parallel bladeenc today. Apparently, his company (www.scyld.com) is releasing their own MPI soon. He suggested that I add a two-line fix to parallel bladeenc that allows MPI_INIT to fail, and then allow it to procede in a serial fashion. This is a truly cool idea, actually. He was motivated by the fact that they support a "serial" MPI dynamic library that allows mpirun-less invocations of MPI programs. In contains stubs of all the MPI functions and simply fails (i.e., returns != MPI_SUCCESS) if you invoke any of them (e.g., MPI_INIT). Hence, if your code is smart, it takes the failure of MPI_INIT to mean that it should run in serial. So I made the quick change to parallel bladeenc; it'll go out in the next release (whenever that is).
Speaking of parallel bladeenc, I mailed Tord about a week or two ago asking questions about the MP3 format itself -- Jeremy Faller and I spent about half a day trying to make parallel bladeenc generate diffable output to serial bladeenc. We didn't succeed, and actually came up with many more questions than answers/solutions, but we understood why parallel bladeenc's output is different than serial bladeenc's. The parallel output is actually probably lower quality -- something we'd like to fix. But we can't do that until we understand the output format of MP3 more... Still waiting for an answer from Tord. :-(