Meeting 126 – 25th April 2019
Stuart, Chris P, Brian R. Sylvie, Rolf, Alex
Andy G. Louise, George on leave.
Stuart: could we replace Hangouts with BlueJeans / Zoom stuff, and our VC stuff?
Rolf: will switch to HZB’s videoconf rooms – pressure not to use Hangouts (too)!
Alex: would need to book VC room, but could do. Stuart says BlueJeans / Zoom can be used on a laptop.
ICAT Server (Stuart):
- Nothing to report
Topcat new snapshot (yesterday):
- console warning if wait for plugins starts to get long
- Add optional preamble HTML to login page, defined in lang.json (issue #415)
- Add showAsButton to authenticator configuration (in topcat.json) to give an authenticator its own login button (issue #418)
- Add extraLoginButtons to facility configuration (in topcat.json) (issue #419)
- Remove DOI and some login redirections from browser history (issue #424)
- Replace hard-wired entity inclusions in metatabs with calculated inclusions (issue #405)
The login/authenticator configuration changes were motivated by ISIS requirements, largely concerning users coming to Topcat from DOI landing pages, who were finding the login page confusing; so more text needed, and separate buttons for anonymous login and for registration with the ISIS User Office); also, redirections within Topcat were making it impossible for them to get back to the DOI landing page by using their browser’s Back button.
IDS Server (Rolf):
- Last week released plugin and storage-file; have discussed changes additional log call main storage; have sent instructions for plugin authors.
- New call only from upcoming version of ids server; so only update plugin then.
- Mostly ready to release ids server; debate today, release tomorrow.
Stuart summarised the issue: behaviour of current IDS – when restoring dataset level, if a file is missing, can cause an exception in the plugin. Rolf would like IDS to throw an exception; but Alex relies on current behaviour where it continues to restore data from the zipfile. Stuart suggests a compromise of making the strict behaviour the default but configurable. Rolf: this may not work. If ICAT is not consistent with storage, IDS will not work. Many other places where this can happen. Risk of downloading a corrupt zipfile.
Stuart: how does it work for Alex at present?
Rolf: it will restore with ids 1.9.0 but zipfile will be corrupt. Alex: impression was it was working fine; but it doesn’t. Test yesterday produced an error. So it may never have worked [after all].
Rolf: new behaviour would throw an error, so DSRestorer would stop.
Stuart: is this reported back to the user?
Rolf: nothing happens: the file does not restore, and so [sits at Restoring…?]
Alex would like to test again. Consider adding [expected] location as new datafile parameter, but will have to be added to every datafile.
Stuart: suggest Alex tests again tomorrow, if still fails, let Rolf continue with his pull request, as it won’t make any difference to what Alex sees.
Rolf: consider visit Grenoble to discuss further.
Stuart: nothing from ISIS (via George)
Chris: no real changes in last few weeks. Interested in progressing discussions on IDS improvements, including new hardware.
Stuart: Rolf suggested using hard links, and that looks like a good solution – one-line change to code, so trying it. Considering a larger change to the tidier process, perhaps use it to point to the area where we create these links, and use it to clean them up. More thought needed.
Rolf: can’t use current tidier to do this.
Stuart: ? separate tidier to manage links? Delete old files, or delete files until some threshold is reached.
Rolf: upgraded production to 4.10, and it works. Also tried using schema changes, successfully.
Chris: DLS still to decide 4.9.3 vs. 4.10 – may not need schema changes.