xls format give two results about fido

[anjackson]

I believe this is the correct result, as PRONOM/DROID/Fido cannot distinguish between these two formats. Your document could be Excel 97 or Excel 2000-2003.

[adamfarquhar]

I guess that the signature ‘does not’ distinguish between the two formats. Perhaps someone could look at improving the signature for this case?

From: Andy Jackson [mailto:[email protected]]
Sent: 06 September 2013 15:34
To: openplanets/fido
Subject: Re: [fido] xls format give two results (#37)

I believe this is the correct result, as PRONOM/DROID/Fido cannot distinguish between these two formats. Your document could be Excel 97 or Excel 2000-2003.

—
Reply to this email directly or view it on GitHub #37 (comment) .

[techmaurice]

Will look into this. If there is a way to advance the signature going to submit this to PRONOM.

[anjackson]

According to this relevant StackOverflow question, it looks like being a rather difficult case, similar in nature to the issues around the TIFF tiff. The base file format is the same for all (BIFF8), and in the same way as for TIF, the only way to determine the particular version of the format is to fully parse the file and look for data that indicates the use of features added by later versions. In some cases, this is manageable via binary signatures (e.g. PNG versions are differentiated by feature usage).

Using a hierarchical format structure can avoid this issue, by having a single BIFF8 format which has two known sub-formats. As PRONOM is not hierarchical, the only options are to move to a new record for BIFF8 and deprecate the others (a la TIFF), or to invest a lot of energy in doing feature-based format distinction within the BIFF8 payload.

[techmaurice]

Thanks Andy.

Feature based identification would be beyond the scope of Fido IMHO.
As for PRONOM not being hierarchical, it might be time for a revision of it. That would take time and money, but I think it is well worth it.

Another option of course is to urge institutions that have need for a broader and more actively maintained registry to finally do something about the the format registry issue. For what I know about former initiatives regarding GDFR and UDFR, it stalls because of the 'tomato, tomatoe' type of discussions about naming, semantics and who would be responsible.

[anjackson]

Back at the DROID 7 consultation event, TNA seemed pretty unlikely to change PRONOM significantly. It's hooked into SDB as well as DROID and I think it's quite expensive to modify. However, that was a while ago now.

FWIW, I feel that the issues behind the failures of UDFR, CRISP and to a lesser extent, Just Solve The Problem are not really about technology or modeling. It simply isn't clear to me that we found the people who will spend time adding the info and then consulting them, prototyping with them, building with them.

I am willing to spend a few hours every week or two on the Archive Team wiki, and I do, because I consider it a part of my job. Who else thinks the same? I don't know.

As for hierarchial format identifiers, we already have them in the form of mime types plus the mime-info specification. Just add a version parameter and you can cross-walk to PRONOM if you wish. Tika supports this approach, is simple enough to extend, and can also be used for feature detection if necessary, so most of my effort goes there right now. Fido is also essentially mime-info based, so it might be possible to patch the hierarchy on top of the PRONOM data, I guess.

[carlwilson]

Closed due lack of recent activity.