> We are obviously seeing things differently. I am not talking about algorithms > that plug into TULIP with *no* modification whatsoever. (That would be where we > need PyExpect.) I am assuming (1) most algorithms run through once, (2) they No need for Expect; you can simply use a bidirectional pipe. Expect becomes necessary only for programs that require interactive input in a way that is not compatible with stdio. I could well imagine that we won't need Expect at all! > communicate via stdin/stdout, and (3) the source code is available and can be > modified by us or by the author. And considering this is the situation, these If we have the code and are allowed to modify it, then we can just as well integratre it into Python properly, without any communications overhead. Writing and parsing XML is slow! > It seems you and Thomas are considering tools not only with various languages > but with various GUI's, that can be added to TULIP without modification. Unless Yes, as a worst case. Maybe we should keep the discussion more pragmatic and consider specific examples for each case; if we can't find an example, we might just as well forget about it... > Well, I don't work with the most common tools either. I can take a quess at > what might be nice...some tools that allow someone to see a 3D structure, > manipulate it, and point out some residues maybe? Anything else? Interactive manipulation is the hardest part... Lots of GUI programming... Konrad. -- ------------------------------------------------------------------------------- Konrad Hinsen | E-Mail: hinsen at cnrs-orleans.fr Centre de Biophysique Moleculaire (CNRS) | Tel.: +33-2.38.25.55.69 Rue Charles Sadron | Fax: +33-2.38.63.15.17 45071 Orleans Cedex 2 | Deutsch/Esperanto/English/ France | Nederlands/Francais -------------------------------------------------------------------------------