Search problem

[PeterJ]

How did you get on with the link I posted earlier [mention]Rorthron[/mention]?

https://spectrumcomputing.co.uk/index.p ... el_id=5522

[Rorthron]

This is tricky. If searches were restricted to a single company, then searching for "Elite Systems" wouldn't give you Batty, since it was published as "Hit-Pak". Searching for "Zigurat Software" wouldn't give you Sir Fred since it was published as "Made in Spain".

In those cases I personally would not want to see either title returned.

For instance, searching for "Grandslam" at ZX-Info has a different behavior despite using exactly the same database:

https://zxinfo.dk/search/grandslam

This is great! I also really like the compact presentation. Sorry, [mention]PeterJ[/mention] , but I'm jumping ship!

([mention]kolbeck[/mention] On a seoarate subject, I notice ZXInfo has the ZXDB "author" problem that was discussed here before. Steinar Lund is not an "author" of The Hunt for Red October. He just did the art for the packaging.)

How did you get on with the link I posted earlier @Rorthron?

https://spectrumcomputing.co.uk/index.p ... el_id=5522

It's better, but still includes some A&F, ASP, Argus Press titles. It's not so many that I can't ignore them, but ZXInfo has the exact information I am looking for.

Perhaps can be more clear if the publisher company has also the owner, ex: Bug-Byte Software (Grandslam Interactive Ltd)? Both as individual links.

Yes, something like that would be better. For example, search could offer a choice of searching by publisher or by group of publishers.

[kolbeck]

...
(@kolbeck On a seoarate subject, I notice ZXInfo has the ZXDB "author" problem that was discussed here before. Steinar Lund is not an "author" of The Hunt for Red October. He just did the art for the packaging.)

Yeah, looks like I missed the conversation about Author/Contributor issue. Will look into that shortly

/Thomas

[Einar Saukas]

Yes, something like that would be better. For example, search could offer a choice of searching by publisher or by group of publishers.

Actually it should be easy to provide a "strict search" option in Advanced Searches. The implementation just need to use "publishers INNER JOIN search_by_names" instead of "search_by_publishers INNER JOIN search_by_names".

I can help with the technical details if anyone's interested

[PeterJ]

Yes, something like that would be better. For example, search could offer a choice of searching by publisher or by group of publishers.

Actually it should be easy to provide a "strict search" option in Advanced Searches. The implementation just need to use "publishers INNER JOIN search_by_names" instead of "search_by_publishers INNER JOIN search_by_names".

I can help with the technical details if anyone's interested

Thank you [mention]Einar Saukas[/mention],

I think that idea makes sense, but before making any changes I would like to gauge other opinions. I don't want to make such a big change only for a few months later someone to complain that the search is to restrictive and they can't find game Y from publisher X.

[PeterJ]

This is great! I also really like the compact presentation. Sorry, @PeterJ , but I'm jumping ship!

ZXInfo is excellent. That is the joy of [mention]Einar Saukas[/mention] making ZXDB opensource. Multiple websites and other services can use the same data but put their own spin on the layout. The two sites share images with each other and [mention]kolbeck[/mention] also contributes very many screens and other materials. Its a perfect partnership!

If you want a really compact view on search you can click 'compact' on the search results filter:

[StooB]

Actually it should be easy to provide a "strict search" option in Advanced Searches. The implementation just need to use "publishers INNER JOIN search_by_names" instead of "search_by_publishers INNER JOIN search_by_names".

I can help with the technical details if anyone's interested

Thank you @Einar Saukas,

I think that idea makes sense, but before making any changes I would like to gauge other opinions. I don't want to make such a big change only for a few months later someone to complain that the search is to restrictive and they can't find game Y from publisher X.

Neither a broad search or a strict search is actually going to produce the "correct" result because the data in ZXDB isn't accurate enough. A strict search for "Grandslam" as publisher would only return titles published by Grandslam on its main label, ignoring titles it published on the Quicksilva, Bug Byte and Bug Byte Premier labels. A broad search goes to the other extreme by including titles like Manic Miner (1983) - which was never published by Grandslam who didn't even exist until 1987 - because they ended up as the owners of Bug Byte.

The problem is that the data on relations between labels inherited from WoS is frozen in time at some point in the 1990's. As far as ZXDB is concerned, Grandslam always owned Bug Byte and Quicksilva, EMAP always owned Crash, Titus always owned Mastertronic etc.

The only way to really solve the problem is by having joint publishers as is already the case for Pac-land which is published by Grandslam and Quicksilva. Without that, it's never going to be possible to do accurate queries such as "all titles actually published by Grandslam" or "Firebird titles published by Microprose rather than BT".

[PeterJ]

The only way I can think of doing what you say is recording the start and end date for each publisher then asking the user in what time in history they want to search for, or having 1 to many relationships all over the place to display the correct information at a point in history. It almost becomes a three dimensional database model.

If you have experience of relational databases you are welcome to plan this out, collate the information, and see if Einar can incorporate it.

[Rorthron]

Neither a broad search or a strict search is actually going to produce the "correct" result because the data in ZXDB isn't accurate enough. A strict search for "Grandslam" as publisher would only return titles published by Grandslam on its main label, ignoring titles it published on the Quicksilva, Bug Byte and Bug Byte Premier labels. A broad search goes to the other extreme by including titles like Manic Miner (1983) - which was never published by Grandslam who didn't even exist until 1987 - because they ended up as the owners of Bug Byte.

I don't think there is actually a problem with either type of search. Both can return "correct" results. Narrow search is certainly correct, as it makes no assumptions about ownership of labels and just uses the labels titles were published under. It is never incorrect to say a Bug-Byte title is a Bug-Byte title.

Broad search as currently implemented does theoretically hit the problem you highlight about changing ownership. It could in theory be incorrect to say a Bug-Byte title is a Grandslam title. But in practice it should make little difference. It is likely Grandslam acquired the rights to Bug-Byte's back catalogue, and the same probably applies for other acquired publishers. Acquisitions of Spectrum software houses since the 90s are pretty much irrelevant. Do we really want to replace Sinclair with BSkyB or Ultimate with Microsoft?

So a narrow search is always correct in its own terms. A broad search isn't always, but it will only rarely be incorrect. It isn't really necessary to track chronological data for labels, though I understand why you might want to add it.

[Einar Saukas]

The only way I can think of doing what you say is recording the start and end date for each publisher then asking the user in what time in history they want to search for

I don't think this would be enough. If company A was created in 1985, then acquired company B in 1987, should we display a game published by B in 1986 when searching for A?

What if B was still selling it in 1987?

I have no objections about storing dates in ZXDB to indicate when each company was bought by another. But even using this information, I don't see an efficient way to "solve" the kind of search some people expect.

[PeterJ]

We will certainly need a choice [mention]Rorthron[/mention], because what's right for you may not be for others. Searching is a very personal thing. It's trying to cover all bases and not restrict the search too much.

I'm sure people are paid huge amounts to work this stuff out!

[PeterJ]

Agreed [mention]Einar Saukas[/mention]. I can't start to imagine the maintenance on that sort of size database.

[druellan]

I keep thinking the problem is not the search but to communicate the user what he/she is searching, and perhaps also providing some choices, like:



IMO this way you not only have an idea what you're searching, but also you can immediately narrow down your query.

[Einar Saukas]

Authors could also have this kind of choice.

Or perhaps a single flag to choose between "broad search mode" and "strict search mode"? This way you don't have to overload the search interface with too many options.

[AndyC]

I'm sure people are paid huge amounts to work this stuff out!

I wish.

The narrow/strict option is certainly possible, but another solution might be to "do a Google" and instead weight the order of results somewhat. Such that everything comes out but things directly released by GrandSlam all come first and then other related publishers follow. That tends to cover both bases since you can easily crop the results of you are only interested in specifics.

[StooB]

It is never incorrect to say a Bug-Byte title is a Bug-Byte title.

There are Bug Byte collectors who would disagree with that - they don't consider anything published by Argus or Grandslam as a "proper" Bug Byte release. It's the same for Quicksilva, Imagine etc.

It could in theory be incorrect to say a Bug-Byte title is a Grandslam title. But in practice it should make little difference. It is likely Grandslam acquired the rights to Bug-Byte's back catalogue, and the same probably applies for other acquired publishers.

You can't make this kind a blanket assumption, and it's not a "theory" either! Like Imagine, Bug Byte was a dead company whose assets were sold by the liquidators. This is a completely different situation to Quicksilva, who were bought by Argus Press as a going concern. In any case, we know that Matthew Smith took Manic Miner to Software Projects, and Bug Byte never actually owned it. In most cases, publishers don't buy the copyright - they licence it for a period of time. If you buy another publisher, you don't automatically get to own everything they ever released.

[PeterJ]

Hi [mention]StooB[/mention], if you are happy to document some examples as to how your proposed structure may look we can certainly review it. Also how you envisage the pages and searching would look. Also is all the required information available somewhere?

The documentation for ZXDB is here.

https://github.com/zxdb/ZXDB/blob/master/README.md

Please feel free to ask any questions about the structure and table links. Sorry I don't know if you are up with database design?

Many thanks.

[Rorthron]

It is never incorrect to say a Bug-Byte title is a Bug-Byte title.

There are Bug Byte collectors who would disagree with that - they don't consider anything published by Argus or Grandslam as a "proper" Bug Byte release. It's the same for Quicksilva, Imagine etc.

There may well be, but that is irrelevant to the point I made. Some users might find it useful to observe the distinction you describe, but they would struggle to show that there isn't an unambiguous definition of Bug-Byte that includes all titles sold under the Bug-Byte label. Your prior contention was that a time-based definition of a publisher was the only possible definition. I am pointing out that there are other valid definitions.

It could in theory be incorrect to say a Bug-Byte title is a Grandslam title. But in practice it should make little difference. It is likely Grandslam acquired the rights to Bug-Byte's back catalogue, and the same probably applies for other acquired publishers.

You can't make this kind a blanket assumption, and it's not a "theory" either! Like Imagine, Bug Byte was a dead company whose assets were sold by the liquidators. This is a completely different situation to Quicksilva, who were bought by Argus Press as a going concern. In any case, we know that Matthew Smith took Manic Miner to Software Projects, and Bug Byte never actually owned it. In most cases, publishers don't buy the copyright - they licence it for a period of time. If you buy another publisher, you don't automatically get to own everything they ever released.

Actually, I did make that assumption, just as you make the assumption that "in most cases, publishers don't buy the copyright". The point is that unless we can identify the exact contractual arrangements for every single game in ZXDB, we have no choice but to make assumptions. It would be great if we had a complete and accurate contractual history for every title, but it is in practice unworkable to get that, so all we have are assumptions. If you have a better set of assumptions, I'd be interested to hear them, but based in the discussion so far, the current assumptions don't seem to me unreasonable, and no-one has suggested a workable alternative.

[StooB]

The point is that unless we can identify the exact contractual arrangements for every single game in ZXDB, we have no choice but to make assumptions. It would be great if we had a complete and accurate contractual history for every title, but it is in practice unworkable to get that, so all we have are assumptions. If you have a better set of assumptions, I'd be interested to hear them, but based in the discussion so far, the current assumptions don't seem to me unreasonable, and no-one has suggested a workable alternative.

Why is there "no choice but to make assumptions"? The database should stick to the actual facts and nothing else. If Manic Miner appears in a list of titles published by Grandslam - which we know is 100% incorrect - then that does seem "unreasonable" to me! What does it say about the credibility of the data if it can't get the details of one of the most famous games on the platform right?

[PeterJ]

The offer is still there to contribute [mention]StooB[/mention].

We want to ensure we make the right choices before investing time in them.

[PeterJ]

If Manic Miner appears in a list of titles published by Grandslam - which we know is 100% incorrect - then that does seem "unreasonable" to me! What does it say about the credibility of the data if it can't get the details of one of the most famous games on the platform right?

I think maybe this needs clarification. Let us be clear that the search results and page detail does say Bug-Byte. The discussion here is about the search results being date sensitive and if to include titles where the company has subsequently been adsorbed at said date. If people want huge changes like this they need to step up and contribute.

[Rorthron]

Why is there "no choice but to make assumptions"?

Because we don't have the complete contractual history of every Spectrum game.

The database should stick to the actual facts and nothing else. If Manic Miner appears in a list of titles published by Grandslam - which we know is 100% incorrect - then that does seem "unreasonable" to me! What does it say about the credibility of the data if it can't get the details of one of the most famous games on the platform right?

Manic Miner is a great illustration of the problem. Not only is it one of the most famous Spectrum games, it is also one of the very few whose contractual arrangements have been openly discussed. Yet we actually know very little about its contractual arrangements. Who currently owns or has distribution rights to Manic Miner? I think Steve Wilcox will claim he does, but I doubt he could back it up. Even for the period where we do have information on Manic Miner, we have little more than the word of a young Matthew Smith, who seems not exactly to have been on top of these things.

In any case, this is one of the most clear-cut cases. What do we know about the contractual arrangements for Program Pack 4?

If you are able to establish all of this information reliably for every Spectrum title, then great, but it seems an impossible ask to me.

[PeterJ]

The advanced search section on the home page now includes an option to specify results strictness. The default behaviour is to include everything, but choose 'strict search' from the pull-down menu and that is what you should get:



I'm aware that the search area on desktop layout needs a bit of design work after adding this extra filter.

[Rorthron]

Seems to work nicely. Thanks!