49
edits
Changes
Some feedback morgamic asked me to contribute
Not sure about:
* How much we need to break apart the frames table?
** ''What you have here looks about right, imho.'' - [[User:Crowder|Crowder]]
* How normalized do we want things?
** ''Err on the side of more rather than less.'' - [[User:Crowder|Crowder]]
* For id->name pairings is it worth pulling the values out since data will be duplicated heavily for things like product, platform, os, etc.? To reduce the size of the main reports table I think it's necessary but I'm open to being convinced otherwise.
** ''For things where you can save time and space, normalize here, otherwise don't.'' - [[User:Crowder|Crowder]]
* Do comment and reason need to be separated from the report table as well?
** ''No'' - [[User:Crowder|Crowder]]
* Need some additional tables not talked about
** Crash groupings by address, module, function?
** For each of these groupings -- bugs open, bugs closed table
** Stored bugzilla queries in each of these groupings pointing to a query that brings up all related bugs
** A table (or set of tables) recording all the DLLs associated and captured with each report (perhaps DLL ID/version, then normalize DLL names to elsewhere) [[User:Crowder|Crowder]]
PostgreSQL 8.2 or higher should be used because it offers things that MySQL does not have:
* Normalize, come up with a snowflake-like schema, abstract up-front but better long-term
* Leave denormalized, just swallow inefficiencies in the short term and pay the price later on
** ''As mentioned above, I vote for the former.'' - [[User:Crowder|Crowder]]
How we can have both:
A major concern about over-normalization is data insertion and the problems it presents for the processor, which should probably be transactional and may have to do lookups and some db magic in order to successfully insert new records.
* ''Should be able to cache repetitive stuff like OS in whatever tool is talking to the DB to save it some work, if necessary'' - [[User:Crowder|Crowder]]
The lookups would be on stuff like product name, build, platform, os, locale, url, etc. I am interested in getting all this stuff out of the main table mainly because there is going to be a ridiculous amount of duplication.
