[Edserplo_users] Update - Edteva stuff

[Steve Loch]

We've resolved problems in relation to statistics and model residual output; also gap filling. These items were preventing the replacement of Edteva by Edserplo. There are in fact still discrepancies in statistics but they are in the MSL count of days, are extremely rare and are down to differences in rounding as between the Java computation and the Fortran one. There also appears to be something wrong with the way Edteva handles Devonport's history. Thus Dataring can be processed purely using Edserplo and this I suggest is now done. Liz says she wants to assess whether she can do all the WOCE work using Edserplo (and Java version of Duxmod I suspect) so I await that. 

I also need to check out the performance of tidal analysis before I power down BISCX. Thanks are due to Jonathan for providing a "history" comparator which allowed me to finish assessing the problems quickly. For those who want to check out the model residual aspects: generate the output once via Edserplo and once via Edteva in different directories (using say rodump/working.dmp) to the same interval - which has to be hours. Then use 

diff -r dir1 dir2 

Take care with the flag transmutation and don't sample.

The NFS monitoring implementation for Edserplo wasn't correct. Hopefully this update will improve matters. We have seen, incidentally,  the occasional discrepancy but most are masked, I suspect, by errors in the monitoring software. Capturing before and after file images without using a lot space is an interesting problem - to which we might have a solution.

At other times Jonathan has been improving lots of things which require an indepth knowledge of Java Swing - like message windows and the banishment of orphan cycle windows. He is currently working on Edserplo usage logging. I will continue on customisation issues (database, production version of obfuscation). The web tooltips service is also part of customisation and awaits some input from Geoff.

Steve