EMBOSS: Project Meeting (Mon 30th October 2006)
The ajdom library is performing well so far. One issue is that as DOM libraries need to store all information in memory we have to consider better ways to extend large numbers of strings. Peter suggested modifying the string code to double string size when appending.This is already done when using formatted writes to a string, but for simple string appends they are extended by only a few bytes each time (typically 32).
Peter is continuing to test the graph title modifications.
Mahmut has updated the EMBRACE Wiki internal pages for all SoapLab services, and there are now 140 documented test cases for SoapLab service operation.
Mahmut has started working on typed SoapLab services using Peter Ernst's SoapLab code, modified to use a single sequence type. This is now working on the test server. Using one sequence type should work well.
Alan raised the question of WS-I web service metadata documentation. Shaun will look into this. Peter will raise the issue at the new EMBRACE executive board meeting and work with the EMBRACE technology work package members. Rodrigo has similar issues with the REST services which have the same basic WSDL as the external services (and EMBRACE) webservices but use different methods. An example if the lack of a clear definition of the output style from RESTful services (XML, HTML, raw text) where service invocation is limited to URL syntax. Maintaining metadata and WSDL files independently of the application provider is especially difficult.
Jon has also written a DocBook guide with a list of tags that needs to be reviewed by the other authors. Only about 20% of the DocBook tags are used. Some of these are mandatory, others optional. Content text is now being added, mainly by adding markup to existing text and checking for consistency of style.
Peter will help by generating text from ACD, application, data structure and function documentation with DocBook markup. Many of the new required tags already exist (e.g. short documentation strings for functions) but are not yet fully populated. We can generate a lot of them from existing descriptions, and extend existing documentation for some new tags.
EBI support are looking into an issue with the SRS server which has become limited to 30 entries when returning the text of sequence entries.
All new issues were considered to be resolved.
Sequence entries would need some naming convention. The Protein ID could be used for coding sequences, at least as an alternative name. Something has to be invented for most other features.
The USA syntax would need extending to request sequence beyond the end of a feature. Negative values are already used for offsets from the end. Perhaps explicit positive numbers could be used - but they would need special processing. Most simply this could use new associated qualifiers and new attributes of the AjPSeqin datatype which ACD processing could populate from the USA. Peter will investigate further possibilities.
The next meeting is on Monday 13th November.
It was agreed to keep the same meeting schedule for next year.