A Case of Mistaken Combination

I'm a little confused by why this would be a bug. Neither title has an ISBN entered. Both use the title Italy. Both use the author George Kish. Isn't this a pretty standard instance of "there's no way a computer could figure it out, so this kind of thing has to be fixed by hand"?

ETA: I know, you say you don't think the system is broken. But I'm just trying to figure out how you would expect this to ever be marked "fixed." It seems like a run-of-the-mill Combiners! problem.

How are they supposed to manually fix this? And even if they could, don't you think they have better things to do than separate out instances where there are two books with the same title, same author, and same ISBN but actually different works (with one copy each, I add, so it's not like it's something many people are going to notice?)

I'm closing this. If you want to open a discussion about cases like this, where there are different works that cannot be disambiguated by the current system architecture, either RSI or Talk About Librarything would be a good place. But it's not a bug.

Exactly. Two books having blank ISBNs is equivalent to them having the same ISBN.

5>

Neither of them have an ISBN.

Yes, that's what I said. They have the same ISBN. It's null in both cases.

3> Or were you recently appointed Sole Arbiter of Bugdom?

I was just trying to ask you for more information. I'm not sure why that has to be taken in such a way. Other than the shot at me, the information in the followup was very helpful. With that explanation, it's more obvious that they can't be separated by users. But, as lorax points out, they don't necessarily need to be. Since they have no ISBN included, it's just a case of "bad data", just as if someone has a book with one author and a title of another.

10> And as for the shot at you, this is twice now you've discussed the viability of threads I've started in this forum.

I ask questions; that's what I do. I think it's the same questions lots of people would have from your original post. It sure seem lorax did.

And again, you've included some helpful information in #11. I really wish you'd mentioned that they don't have ISBNs at all up front, but that's how these things go so I don't blame you. I think you just take these things far more personally than perhaps you should. It's not like you never question anyone else's posts. :D

I'm not sure it is a bug, but it should be fixed, and only staff can do it. It affects more than two books, because of the potential combination that is right for one of the books but not the other.

This is not a case of bad data. The books are apparently entered correctly. Nor have these entries been abandoned. Both members are still active.

I think I remember a discussion some time ago about the few cases where one author wrote more than one book with the same title for various series. This happens in guide books and similar, as in this case, but also in cookbooks and in art books about specific artists. I don't think there can be something programmed, but discussion and a policy to deal with these cases would be important.

#11

I'm confused here.

I find lots of potential combinations on the editions page that should never be combined.

See this page:

http://www.librarything.com/work/2793952/editions

I don't think anyone would combine them but they are there. I don't think of this as a bug.

I think I chose my words poorly when I said "bad data." What I was trying to say is "incomplete data." Like when someone puts the name of a book but no author. There could be two different authors who wrote a book of the same name.

I was also not trying to say you couldn't infer what the user wanted by looking at the OTHER data in their catalog. I was just coming from the standpoint of how the system works without manual intervention from Combiners! (amongst others). It doesn't see any data other than title, author and ISBN, right? That's why I was asking about how the system would tell them apart. It's an honest question, not rhetorical. Can LT tell apart two works that have the exact same author and title and have no ISBN? Will it always constantly be trying to combine them back together?

I don't see that this is a bug, though I suppose that depends on your definition.

To me, a bug is something that does not work as it is intended to work. The system is designed to automatically combine works with the same title, author & ISBN (or lack thereof). It did so. So it's not a bug.

Would it be a good thing if the system could take into account additional available data in making these combinations? Sure. That's a "recommended site improvement".

18> Glad we could kiss and make up. :D This is actually what I was trying to get at in my first response to your post! I think sometimes it's hard to tell a rhetorical question from an honest one. My question in post #2 was definitely an honest one.

#16

The relevance is that there are more examples than the one you just gave. It adds to your argument as your example is not the only one.

8>

Now you're being deliberately obtuse.

All books with no ISBN have the same ISBN.

All books that have the same author, title, and ISBN are considered to be the same by LT. It doesn't matter whether the title, and ISBN are "Programming Collective Intelligence" by Toby Segaran, ISBN 0596529325 (which is here on my desk) or "Italy" by George Kish, ISBN NULL; in each case, all three slots match, so they're considered a single work. Does this misfire in some cases, especially for ISBN-less works? Sure, which is why I suggested posting an RSI. Is this a bug? No.

#22

Your example for the work and the editions page of the other work show combinations that aren't correct - same series. The Maclaren shows the different volumes in three different compilations of the work. Also not correct.

The reason I asked the question (and the confusion) is I see this a lot. Should it be a concern?

I'm a new member (about 6 months) and still learning.

22> I think it's just a case of confused expectations. If your subject line had been "Can we get some staff member to separate these two editions? - it won't let me." I think it would have clicked with us all up front. It's not a typical request, and there really isn't any ONE good place for it. I could see this as a post to Talk about Librarything, Combiners! (though less likely to get staff attention), a profile message to Tim (though that'd just be Tim, not all staff), etc.

I think the New Bug Collectors (now with Bug Reports!) set up expectations in a different way. When I read something here, I don't expect it to be a request for the staff to just go tweak something in the data. That causes me to "read between the wrong lines", since I'm expecting we're talking about a different subject.

But I can see how once I understand what you're wanting, it's a perfectly valid request! :)

#25

Ah! I just got it by looking at the covers for Kish's books.

Sorry to be so dense but I now understand what you are describing.

Thanks for your patience.

22>

Ah, so it's not the general case of the autocombiner misfiring that you're concerned about, but just this particular case? Sorry I misunderstood.

What I was trying to get at earlier, among other things, is that these can't be separated. I mean, maybe the staff could go in and manually set the work code for one of them, but there's no way to do it working within the system, and I think it's a bad precedent to have them spend a lot of time on this sort of thing (especially because the books could be recombined later.) Leaving a disambiguation notice, and in this case (since there's only one copy of each book) contacting the owners of the books, would seem to be a more appropriate and more sustainable course of action.

I was curious as to what it might do if I tried to separate the only edition from itself (on the off chance it would split them into two works). Tried it and promptly got a trained monkeys error. It left behind a 0 copy work and separated them both out together into a new 2 copy work. So I recombined that back to the now 0 copy work.

After recombination, it now has 3 copies.

Here's the new one:

http://www.librarything.com/work/10531289/details/65698099

Just can't win. :D

ETA: Oh, and another interesting note. The separation that got the monkeys error did NOT show up in the separation log. Interesting, that.

Fine by me. I'd just like to stress that I combined the new 2 copy work with the 0 copy one it had left, thus ending up with a 3 copy work. It sucked in another work all on its own. Oh, and I didn't actually do anything with the disambiguation notice. LT did that all on its own.

The information on the disambiguation notice would certainly be ambiguous now. I was trying to separate the works by the number of pages, the work with 62 pages should not be combined with the work with 160 pages etc. There is another work btw, which has 125 p., I just added it to one of my other accounts.

http://www.librarything.com/work/10531289/details/65698099>

I entered this using OverCat, so LibraryThing should have more of the details about the book. I was wondering if that might make a difference in the automatic combining, since at least one of the other works was entered from a library source too. It didn't. This one too combined with the other.

I suggest we could write a disambiguation note containing the LOC Permalinks for each of the work entries. Do you think that might be enough information to distinguish between the works?

An edition in LT is a unique combination of Title, Author, and ISBN. If two or more books share these values they *are* the same edition as far as LT is concerned. No one, not even Tim, can separate them.

I would consider this a bug in that two books that are different are combined and cannot be separated. Unfortunately, it's one that would take a complete rewrite of LT to fix and so its not going to get fixed.

Easy way to fix:
A book owner adds a fake ISBN to their copy (9999999999) or changes the title slightly (add series info?).
The copy with the fake ISBN or different title is separated.

33>

But that's what we keep trying to tell you; they can't be separated.

37> No they can't. Not without completely redesigning LT.

ETA: It's a basic part of the design of LT. I guess Tim never dreamed that there could be two different books with the same Title and Author (and ISBN if there is one).

37>

That's what I said, and you chose to ignore, back in #22. Sure, the staff could manually reset the work code. But they have better things to do. They can't separate these easily or within the usual workings of the system, and I'd really rather see them working on fixing bugs or new feature development rather than diving into the guts of the system to separate two obscure single-copy works that are just going to be recombined anyway the next time someone sees them.

39> Sure, the staff could manually reset the work code.

No, I really don't think that would help.

What would be needed, for this to be separable, is to have editions distinguished by more than title + author + ISBN: to include publication date and/or publication details as fields that can also distinguish editions. Right now things aren't set up that way. Maybe they could be.

Thank you rsterling, that would help. Tim has been talking about adding more publications details, even if the book description were added to the Detail pages, and if somehow we could separate the editions manually based on the details if they indicate more than one work....at the moment, without more descriptive information we can't really know that this work is really three different works.

http://www.librarything.com/work/10531289/details/65698099
http://www.librarything.com/work/10531289/details/50200328
http://www.librarything.com/work/10531289/details/41986830

Okay, I'm going to be going back through some of the much-older bug reports today and trying to figure out if they're still issues (I'm also going to be trying to hunt down and close duplicate reports). This one, I think, is one of those that's been overtaken by other changes, so I'm going to close it. If the root issue remains, please file a new report with a relevant example.

The root issue remains but I think it's more of an RSI than a bug. LT defines editions by Title/Author/ISBN. Occasionally two books are identical across those three fields but are, in fact, different works. Sometimes this is a result of ratty data, for instance someone entering a movie but with the ISBN for the book. In the case at the top* there are two different works by George Kish, both titled Italy and both predating ISBNs. They are inseparably part of the same LT edition. Fixing this problem would require a new LT feature.

*The links in Collectorator's post are no longer valid. Here are the current links:
Work: http://www.librarything.com/work/2399145
Around the World edition: http://www.librarything.com/work/2399145/details/41986830
Life in Europe edition: http://www.librarything.com/work/2399145/details/50200328

#45 by @jbd1> I really don't think it has, Jeremy. It's just a result of the design of the LT concept of an "edition". An edition is a combination of author, title and ISBN. If two books have the same author and title and were never given an ISBN, they can't be kept separate. It's just how LT works.

ETA: Dang it, eromsted! ;)

Right, I see. I'll keep this example handy for when we dig into versions/editions/&c. again.

ETA: Ditto http://www.librarything.com/topic/139134