Graphics/sound support

[erkyrath]

The JSON graphics commands have already been specced out and implemented in GlkOte. We would need analogous sound commands too.

Gestalt handling is a little weird. In principle RemGlk cannot know whether its display partner supports that stuff. (Sleepmask/Floyd is an obvious case that cannot!) So I think we'll want command-line options for RemGlk to say "graphics, sound, timer support: yes/no". That will determine the gestalt responses, and also whether RemGlk bothers to kick out those JSON commands.

[BroadcastGames]

Perhaps HyperLinks too. Unless I compiled it wrong or something, right now HyperLinks crash remote-if.py if the game has them. For example, CityofSecrets_v3 on Git interpreter if you type the 'about' command the webserver crashes.

EDIT: I see that running with RemGlk param "-stderr" keeps it from crashing on encountering a HyperLink.

With "-stderr" on, the CityOfSecrets v3 menu on the bottom actually work pretty well on GlkOte. You have to fiddle the cursor to a blank spot. For example, start the story and do "z", "z", "z" and a Dark Lady will ask you a question which you can pick prompts. I was able to get my cursor into that window and use the arrow up/down and number key. Anyway, figured that positive experience was worth sharing.

[erkyrath]

Formally, this is a bug in CityofSecrets_v3. It should not make hyperlink requests on an interpreter that does not support them.

However, you're right that hyperlink support should be added. That's the easy one.

[erkyrath]

Timer support is in, and works correctly with glkote amd regtest.py.

[curiousdannii]

Could GlkOte send a list of what it supports in its 'init' message? I'm looking at using glkapi.js in node and I think the only think it has which won't work in node is the has_canvas check. But if remglk could also get its information from glkote then that would be simpler than having to specify several command line arguments. (Though they could be kept to overwrite it I guess.)

GlkOte should probably also send its version number in its init message so that remote systems can check it.

[erkyrath]

Yes, the support list is a good idea.

Not sure about the version number. I generally like version transparency, but RemGlk can't do anything with that information.

[erkyrath]

...now that I think about it, it might not even be GlkOte on the other end of the line. (Regtest.py and Sleepmask use the same protocol to talk to RemGlk.) So relying on a version number for anything is dangerous.

[curiousdannii]

Maybe a version number for the protocol then? I was thinking of situations like the timer function where a VM might want to hand it off to the remote IO implementation but be unable to tell whether that is supported. Of course now we can tell because of the support array, but something else like that might arise again. Or when such a problem arises we can deal with it then.

[erkyrath]

The support array seems like the right way to handle this. It covers RemGlk's needs, anyhow.

[erkyrath]

Hyperlink support implemented.

[BroadcastGames]

On graphics and sound, is there anticipation that RemGlk will be responsible to extract graphic and sound payload elements from the blorb file to hand off to middleware? Write binary files to disk in a data directory and send paths - or putting base64 of the elements into the JSON stream?

[erkyrath]

No, I think that image/sound extracting is better handled by the application framework or launcher.

RemGlk will refer to images only by Blorb number. (See http://eblong.com/zarf/glk/glkote/docs.html#graphicsupdate , and also farther down under "Line Data".) Before launch, the display code will want to unpack all the resources. https://github.com/erkyrath/glk-dev/blob/master/blorbtool.py has a sample implementation of this.

[BroadcastGames]

Ok, Thank you for explaining your direction on the asset management.

For testing sake, i think this story has the kind of complex windowing and HyperLink support in a complex-but-understandable way: http://ifdb.tads.org/viewgame?id=q5el4dphbf7q6e5y GlkOte almost runs it (perfectly) currently but I think it must violate the issue of window Input origin when you click the HyperLink because that generates an error. That story is almost 10 years old, I'm trying to scratch together working Inform 7 source code to reproduce that exact layout.

[erkyrath]

There are, unfortunately, a lot of games that make invalid calls which are not reported in the IDE interpreter or Gargoyle. I recommend testing in Lectrote.