Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Jan 31 2019 22:39
    dmitrizagidulin commented #1061
  • Jan 31 2019 22:10
  • Jan 31 2019 22:04
    kidehen commented #1061
  • Jan 31 2019 22:00
    elf-pavlik commented #1061
  • Jan 31 2019 21:49
    kidehen commented #1061
  • Jan 31 2019 21:39
    kjetilk labeled #1081
  • Jan 31 2019 21:38
    kjetilk labeled #1081
  • Jan 31 2019 21:28
    dmitrizagidulin commented #1061
  • Jan 31 2019 20:47
    cblakeley commented #1061
  • Jan 31 2019 18:42
    elf-pavlik edited #1081
  • Jan 31 2019 18:42
    elf-pavlik opened #1081
  • Jan 31 2019 14:20
    joachimvh commented #1080
  • Jan 31 2019 09:27
    kjetilk labeled #1080
  • Jan 31 2019 09:27
    kjetilk assigned #1080
  • Jan 31 2019 09:27
    kjetilk opened #1080
  • Jan 31 2019 02:27
    elf-pavlik opened #1079
  • Jan 30 2019 16:16
  • Jan 30 2019 15:05
    dmitrizagidulin commented #1061
  • Jan 30 2019 14:23
    csarven opened #1078
  • Jan 30 2019 12:54
    cblakeley commented #1061
Tim Berners-Lee
@timbl
If you write code that runs over file:// just as well as htts:// then it won’t work with design A, unless we change how the solid stack implements file://
elf Pavlik
@elf-pavlik
Maybe PUT and PATCH on the container just needs to check if it doesn't affect statements with ldp:contains in unintended way?
Tim Berners-Lee
@timbl
Like it should filter out for any lip:contains triples, yes
elf Pavlik
@elf-pavlik
to get started it can just respond with error if request tries to modify ldp:contains later could get improved
Fred Gibson
@gibsonf1
@timbl Thanks for clarifying - my version of the Solid Server is running as a graph db, so I have harder time thinking about the file implications. I'm a fan of having as few files as possible and as many triples as possible instead.
Tim Berners-Lee
@timbl
or return 409 Conflict
Sarven Capadisli
@csarven
@timbl Just to be clear, #547 didn't remove it. The issue was marked as revisit and then closed. You can still write to /.meta and see the changes in /
Sarven Capadisli
@csarven
@gibsonf1 Although NSS leans on the filesystem approach, AFAIK, it was never intended to exclude other storage mechanisms. There was (or paused?) work on having different tiers.
I'm all for getting rid of .meta from the interface, and let the server handle it however it wants to internally.
Tim Berners-Lee
@timbl
So at the moment you can write to .meta and read it in the folder itself?
Sarven Capadisli
@csarven
yes
IIRC, that's not the case with regular resources like (Non)RDFSources. foo.ttl.meta or foo.png.meta would be there as separate resources, but that there is a link from foo.ttl and foo.png to corresponding .meta. Container is unique in that /.meta shows up in /
Tim Berners-Lee
@timbl
Is that documented anywhere?
Sarven Capadisli
@csarven
If we apply the .meta idea, it should be consistent across resources. Or just dump it altogether from the interface. I don't see much use in interacting with .meta to add metadata stuff for foo.ttl.meta or foo.png.meta .. or hardly much anything beyond server-generated stuff like posix or oh yeah?
Can't remember.. /me checks
Scott McWhirter
@konobi
I'd see more use for it in the operational space (so non-public)... for example, providing a list of server identifiers of where there is a resource copy of the blob or a list of regions that the file is currently primarily for
Sarven Capadisli
@csarven
@konobi Any particular reason (or real use cases?) why the resource in question can't be self-describing?
Tim Berners-Lee
@timbl
I think the motivating use case is a non rdf resource like a photo or video
So you could build a photo manager around a raw folder of image files
Sarven Capadisli
@csarven
EXIF comes to mind.
EXIF/RDF rather.. or whatever it was called.
Tim Berners-Lee
@timbl
Or you could just put in your own data files among the images
Scott McWhirter
@konobi
@csarven mmm... I'll have to think on it a little... you're right in that container is an odd one
Sarven Capadisli
@csarven
Would you say that an RDF description of that sort can simply be a different representation of the media file? So, can it be content-negotiated or is it really different - don't mean to get philosophical about ranges here.
So, fetching foo.png with text/turtle should return foo.png.meta payload
Tim Berners-Lee
@timbl
I don’t know of any code that uses it so hence the proposal to remove the feature to keep the platform simple
No @csarven I would definitely not go down the Conneg path here. The picture and the users comment about it are not the same information
Scott McWhirter
@konobi
The only thing popping to mind is wondering how caching would play into the linked description, but I'd really need to read up on the state management of caching
Sarven Capadisli
@csarven
When I asked that I had EXIF-like data in mind, as opposed to user annotations about the media file.
Scary movie time.
Scott McWhirter
@konobi
@csarven I suppose there's a delineation in terms of what you consider the meta of the file, which is probably different based on the perspective. In one case it's the metadata of content or derived from the content in the blob; the other is metadata of the "box" in which the blob is stored (or otherwise implemented) but is unrelated to the content inside the file and likely to be ephemeral.
Sarven Capadisli
@csarven
@timbl I'm not sure if removing an application's ability to update the container IRI makes the platform simple. It punts the problem by setting a constraint ie. some resources can be written, some not. That may be a fine (arbitrary) implementation decision for NSS, but it doesn't feel right for a Solid-server implementation. "Solid" need not have to make a hard call on that based on the notion of containers and resources (which is closely tied to LDP-centric approach right now). If we want the backend to be arranged in different ways, we should remain flexible IMO.
As for the practicality of .meta, I think we agree that we don't want to see that in the interface.
Sarven Capadisli
@csarven
Disallowing updates have other side-effects which may or not be intended. As we discussed, constraining container IRI to read-only would mean that we no longer have labels. Not to forget WebIDs with base ending with / (will anyone thinking of my /#i)? The LDP-centric treatment of containers and resources (mapping to / and /foo on disk or whatever) has that consequence. Getting people to migrate their website to Solid-based server is going to be an issue if they want to publish their homepage from / but either have to resort to what LDP says for the root container, or they have to manually edit /index.html or something.
csarven @csarven goes back to thesis mode.
Tim Berners-Lee
@timbl
I think I am persuaded by your argument that allowing triples in everything is simpler
Fred Gibson
@gibsonf1
That would be excellent!
lin onetwo
@linonetwo
Any recent consideration about solid-server plugin/middleware?
Scott McWhirter
@konobi
node-solid-server is built on express, so middleware is definitely possible
Scott McWhirter
@konobi
looking at the issue and the forum though, it seems like you would like to ensure that the internal apis are available in a consistent way to middleware and exposure via config?
aaaagile
@aaaagile
I have given up trying for a solid Windows 10 server.
lin onetwo
@linonetwo
@konobi Express Yes, but express middleware can not reach the internal processing inside NSS. So hoping NSS can expose more internal lifecycle hooks to the config. In this way, custom DB support will be easier.
And currently, the only way to access data is through OIDC, which only works for the web app. If there were a Personal Access Token system which allows a 3rd-party server-side app to access user data with ACL control, complex app likes cronjob-crawler may become possible.
Scott McWhirter
@konobi
@aaaagile what sort of issues are you running into?
lin onetwo
@linonetwo
I thought that Personal Access Token system can be implement on the top of internal lifecycle hooks, if there were some.
Scott McWhirter
@konobi
@linonetwo I believe there's a desire to do some reworking to enable more options, though I'd hazard a guess that gathering a sense of the use-case patterns would instruct that better
lin onetwo
@linonetwo
I just want to make a cronjob-crawler running on my server that sent content to my POD, how can I do this currently? Maybe I missed some way to implement it that is much simplier.