Hello Alexander, [...] > > I know Marc21 reasonably well, and I don't remember now any case > > where having different indicators mean something so different that > > has to be treat differently. > > Here I would be more careful. Basically, I would treat Marc fields and > indicators not as 3 digits plus two other funny chars but consider the > whole bunch as a 5 character wide filed designation. I think, here I'm > in fact a bit more in line with Estebans approach. At least if I > understand it correctly. (Though I agree with you that one might not > come up with a complete bibfield list, but just with a set of "most > common usages".)
But «most common usages» won't cover them all, and so, you cannot load arbitrary records coming from unknown sources and expect Invenio to do the expected thing with them. So, what I'm proposing is: let «any» be the rule, and let's cover the exceptions later. What I'm proposing is to follow the Postel's rule: Be conservative in what you do, be liberal in what you accept from others. <http://en.wikipedia.org/wiki/Robustness_principle> [...] > >> The JSON structure that we create from Marc21 (or anything else) > >> contains as much information from the master format as you want > >> (even the indicators). Meaning that if your data model is well > >> written it is a lossless conversion and there is a one to one > >> mapping that makes possible doing Marc21 to JSON to Marc21. > > > > I hope that. > > +1 > > I share some concerns about this with Ferran and Martin and some > others, and I'm very sure it's quite a task... I don't think it is so difficult if the code just accepts 245%% for title, 100%% for first author, etc. With a 10% effort we could cover more than 95% of the cases. Alexander, would you accept to exchange the current Invenio default behaviour with the default I'm proposing? Knowing that it would not be perfect, do you think that it would be better? Best regards, Ferran
