Mobyle

===> Mobyle official web page

Page updated on: June, 30 2008.
We (Bertrand Neron, Hervé Ménager, Pierre Tuffery and Catherine Letondal) have started a re-development of Pise which aims to extend it, but also, of course, to fix several issues. I list below several useful suggestions that have been made over the past years, and I also includes various ideas we had to improve the software. The aim of this page is to start a discussion, so do not hesitate to send your suggestions or comments! (see contact information).
  1. Poster and position papers:

  2. Bug fixes:

    1. Pise version management, CVS
    2. various installation problems :
    3. bad distinction of local settings (web.pl, databanks list) and general files
    4. make errors badly handled
    5. under some Linux (some Redhat), the signal mechanims does not work, resulting in result page return whereas the program has not finished
    6. batch system (PBs, LSF,...) setting could be made easier
    7. the sequence format conversion system will be made more efficient thanks to Nicolas Joly conversion program (squizz)
    8. programs not installed should not be listed in piping menus


  3. Improvements:

    1. piping types should be more precise (dna vs protein, alignment vs sequences) and some polymorphism should be included (this is also related to the integration of web services)
    2. program versions are not really taken into account, leading to problems for Pise installers not having the same version as the one we currently have installed at Pasteur
    3. databanks settings should be more flexible:
    4. list of local databanks
    5. properties associated to db in order to know which one has a blast/fasta/paracel/... access, what type of db it is, etc...
    6. the email has always been a problem: although it is now optional, the current design is not flexible enough. Email input should be required depending on more criteria:
      1. for specific programs
      2. only for long jobs
      3. only for external users (?)
      4. etc...
    7. data input: the design of forms could be made much easier by enabling common data input among several forms (this is inpired by SeWer, see http://bioweb.pasteur.fr/SeWeR)
    8. various "Pise portals" have been built: for now, we preferred to let the Pise installer choose how to offer the programs they actually have locally. But a choice of several portal flavours could be submitted by those of the Pise installers that wish do share their ideas? (including me)


  4. Extensions:

    1. better pipeline support: G-pipe, developped by Alexander Garcia and Samuel Thoraval (http://www.pasteur.fr/recherche/unites/sis/Pise/#gpipe) is a great start
    2. we are working on the HCI/GUI aspect of the problem (see: Participatory Design of Pipeline tools and Web services in bioinformatics presented at a Workshop on Requirements Capture for Collaboration in eScience )
    3. integration of Web services through the BioMoby framework (http://www.biomoby.org) (we are currently installing some biomoby services and have started to participate to the BioMoby project by writing some documentation on it): the aim is to have a Pise service not only available through a Web form, but also through a services directory.
    4. wrapping of Web servers (instead of only local Unix programs): biologists rarely find everything they need in a single Web site. This feature would be provided for external Web sites that agree to be wrapped. Wrapping would be manual, not automatic (just like the XML definitions for Unix programs wrapping).
    5. there is a need for a simple local menu system, available from the Unix shell to local users, and generated by Pise definitions.


  5. Software architecture:

    1. current code organisation is as you know very problematic: too much metaprogramming, bad modularity, etc... which makes the software unmaintainable
    2. we plan to change the programming language of Pise (perl -> python) (mainly for object-orientation aspects) (don't worry, python is easy to install)
    3. I believe that a core module for local program execution could be common to several types of interfaces (e.g: CGI, local menus, graphical UI, ...)
    4. modules for results gathering probably differ (just because Web interfaces require a temporary directory, which is not necessarily the case for a local execution that could run in a user's current directory)


  6. Draft DTD

    Main changes include:
    1. wrapping of Web servers
    2. some presentation elements (layout and formfield attributes to enforce a type of field - e.g select / radiobutton)
    3. better design for some elements such as paragraph or elements needing code
    4. some renaming (label->info, group->argpos, ...)
    See: mobyle.dtd

  7. Other suggestions (from the mailing-list):

    1. Could I suggest you add support for clusters - applications like mpiBLAST or parallel clustal and so on. I don't know how easy/appropriate it'd be to add that to PISE, rather than adding PISE to packages like BioBrew.
    2. Multiple databases selection for Blast.
    3. For the pipes, it would be good if there was a mechanism to pass more than one file from a result through the same pipe to another program.
    4. Also, it would be good to be able to run more than one piped program from the same result.