Bug: Upload prevented due to "duplicate" detection

[haydenmuhl]

I'm trying to upload a score to this page.

http://imslp.org/wiki/Mal%C3%ACa_%28Tos ... o_Paolo%29

The PDF I'm trying to upload has the file name "malia.pdf". When I try to upload it, the wiki tells me that it is a duplicate of this file.

http://imslp.org/wiki/File:PMLP116116-malia.pdf

They are completely different files, and the only reason I can think that they would be identified as duplicates is because of the name. I expect I can work around this easily by renaming my file and uploading, but is this the desired behavior of the wiki?

[pml]

Haydenmuhl,

Almost certainly a naming issue. It pays to be original! (hint, hint) I see you’ve resolved the problem by using a capital letter for your filename.

The system is there to prevent someone else from coming along overwriting all of your files, which generally is a good thing.

Cheers, PML

[haydenmuhl]

Yes, we don't want people overwriting each other's files, but the behavior still seems odd to me.

It looks like the system is taking the name of the file being uploaded and adds some prefix. I'm guessing the prefix is an internal ID for the work page, or something to that effect. So, if someone else uploads "song.pdf" and then I later upload a different file called "song.pdf" to the same work page, both will resolve to "prefix1234-song.pdf" and the wiki will think I'm trying to update an existing score.

Why I consider this a bug is that I get this behavior when I click the "Add a score file to this page" link. The name implies adding a new score. If I wanted to update an existing score, I would expect some option to select which existing score I want to update. Relying on the name of the uploaded file to identify the score to update seems... kludgey?

Here are a couple possible fixes I would suggest.
* A new upload naming convention other than using the page ID as a prefix
* When a name collision occurs, allow the user to amend the generated file name

Edit: At the very least, the submission should fail and the work page should not be updated. As it is, you get a new score listing that links to the old score. Take a look at the history of the work page I linked in the OP to see what I'm talking about.

[pml]

The prefix directly relates to the workpage and is as good a way of preventing naming ambiguities as any other - there are 90,000 PDFs in the file system you know!

However it does mean works uploaded to the same workpage share the same prefix, and thus you need to take a little care to make your contribution unique from any others on the page already. The subset of possible duplicate names is thus made much smaller, but not always zero.

As for the error when uploading, I gather it's an extra bell and whistle that should have been programmed into the upload pages, but hasn't been because the problem is very infrequent.

The page number (usually of the form PMLP200500) is just as searchable as the IMSLP numbers.

Replacing files is also straightforward - clicking on the small-type link for the IMSLP # rather than the file description takes you to the file location where you have the option to upload a replacement file.

When I looked at the page the two files were distinct — were you logged in or not?

Cheers Philip

[haydenmuhl]

Yes, the page is fine now. I capitalized the first letter of my file, which resolved the name conflict and uploaded the file without a problem.

But take a look at this version. This was one of my first attempts.

http://imslp.org/index.php?title=Mal%C3 ... did=694805

You'll notice that there are two "different" scores listed, but both point to file number 56215. IMO, this is incorrect behavior, because the change then has to be backed out. If someone is not paying attention, they could think they've uploaded their score, and then leave the page in a messy state.

Simply aborting the upload and making sure the work page is unchanged would probably be the simplest fix, if a bit un-user-friendly. But as you say, this is a rare problem, so it would probably be an adequate fix.

[pml]

Another way of looking at this is that it's a smaller bug arising from squashing a much larger and more serious bug (namely that previously it was possible for anyone to overwrite anybody else's uploads). However, the edit you linked did have enough information to work out what had happened — and your correct title had gone into the copyright tagging system, so even if you hadn’t fixed it, someone would have picked up the discrepancy.