Empty files

[NaumannKai]

DROID identifies empty files (size 0 bytes) with certain extensions as files with a certain format, often giving several results.

Regarding the file's content, this information is quite misleading and may cause converting software to crash. A format is a given structure of a content - no content, no structure, no format, right?

We suggest - unless already suggested and declined - to add a feature in DROID that classifies this file type as "Empty file". Of course, this result cannot be directly achieved through the signature file, I guess, but could be delivered like all other results.

What do you think? The other option would be to keep the (kind of dodgy) DROID result and add an "Empty file" flag to file objects in our repository.

Best regards,
Kai
(big PRONOM fan)

[sparkhi]

Hi Kai,

DROID outputs the size of the file, could that be useful to you to filter the results
I believe, the identification is happening based on the extension of the file as it is not matching any signature. I created a couple of empty files with only the .txt being identified as a text file based on extension.

If I change the preferences to Match extensions against all known extensions then it identifies the empty file (with .txt) extension as 3 formats ("TRIM Context Reference File", "ESRI ArcInfo Coverage Annotation File", "Plain Text File"). In all of these the size of the file is there. Hopefully you can make use of it to filter out the results that you do not want.

Hope this helps

[steve-daly]

Yes, that's an interesting point. I agree that a zero byte file can't really be a text file (for example) even if the extension would suggest that it was.

My main concern with making such a change would be the impact for people with automation system using that field. There might be software only expecting a limited set of values for that field, and not 'Empty file' or something similar.

We could possibly make the behaviour optional (e.g. a setting, or command line switch) but we don't want to overload users with settings.

We will have a think about this. Out of interest, are you using DROID from the command line (or API) or would you be looking for this behaviour in the GUI too? It's easier to introduce special/advanced behaviour in the command line, as we don't have to worry so much about the look and feel.

[NaumannKai]

Hi sparkhi and steve, thank you for your immediate replies and your tests. What sparhki found, having three or more results, often happens with common formats like PDF or DOCX, since there is no content or containers for further reference.

Emptiness looks like a pseudo-format to me, but Operating Systems mostly pretent it is, say, DOCX or whatever.

We might also want to copy this slightly strange but common perception.

But if we follow this ticket, regarding automation, 'Empty file' would have to be assigned a PUID in the Signature file as well. Can the Signature file concept be extended in this regard?

We are using DROID "under the hood" of our suite of preservation system modules, the pre-ingest module, the ingest modules, the storage module. No use of the DROID GUI except for testing.

[Dclipsham]

Empty files are sometimes used as sentinels/markers/process indicators,so e.g. a .lock file to indicate that a file is in use, files that might have an extension .inprogress, .failed, .complete to signify that state of some process, .gitkeep signifies to git to keep an otherwise empty directory under source control, etc

I dont think any of these are currently in pronom (maybe dreamweaver lock file?) But this real-world, purposeful and semantically meaningful use of empty files makes me hesitant to have an 'empty file' catchall puid.

Just my two cents as an outsider :)

David