I have deeply mixed feelings about #ActivityPub's adoption of JSON-LD, as someone who's spent way too long dealing with it while building #Fedify.
-
@sl007 @julian I admit I didn't pay attention to immers at the time - I don't play games, not even chess. I was just using chess as an example, didn't mean to trigger anyone's trauma!
Still, it kinda proves my point. You have to use standard AS vocabulary because Mastodon, and if you squint then sure,
TravelandArrive, why not? But given some of the conversations I've seen on this forum, I shudder to think how that would go down if you tried to get approval for that usage from "the community" first.@mat
Just btw, this is 7 years old https://www.reddit.com/r/chess/comments/94ubnd/chess_over_activitypub/ but anywayHowever, given that I have, including immers and redaktor, at least 3 apps where I can use the first chess spec.:
if more than 2 implementations will also support this second chess specification, I will do so too. -
@hongminhee @jalefkowit huh. Iโve been pondering using it for some projects of mine, so this is good to know.
Is it a fundamental problem with JSON-LD, such that it should just be avoided, or a problem with how ActivityPub uses it?
And is there something else youโd recommend that fulfills the same goals?
@lkanies@hachyderm.io @jalefkowit@vmst.io To be honest, I'm not too sure myself. I just know that JSON-LD was originally planned as a foundation for the Semantic Web. I can only guess that if ontology is useful in a certain area, then JSON-LD would probably be useful there too.
-
@hongminhee do you use the activitystrea.ms module from npm? It takes a lot of the pain out.
-
@hongminhee I agree that new developers should use a JSON-LD processor. It saves a lot of heartache.
-
@hongminhee do you use the activitystrea.ms module from npm? It takes a lot of the pain out.
@evan@cosocial.ca I don't remember exactly, but I think I came across it while doing research before developing Fedify. I probably didn't use it because the TypeScript type definitions were missing. In the end, I ended up making something similar in Fedify anyway.
-
@mariusor @hongminhee The
@contextis not supposed to be required in the first place, but here we are adding it to every activity and wasting bandwidth because Mastodon developers didn't read the spec.@silverpill I'm sorry, I'm not aware of that and I thought I read the specs pretty thoroughly. Could you point me in the right direction for where you got this information from?
-
@hongminhee I have the same feeling. The idea behind JSON-LD is nice, but it isn't widely available, so developing with it becomes a headache: do I want to create a JSON-LD processor, spending twice the time I wanted to, or do I just consider it as JSON for now and hope someone will make a JSON-LD processor soon? Often, the answer is the latter, because it's a big task that we're not looking for when creating fedi software.
-
@silverpill I'm sorry, I'm not aware of that and I thought I read the specs pretty thoroughly. Could you point me in the right direction for where you got this information from?
@contextis a recommendation, not a requirement.ActivityPub:
ActivityPub
The ActivityPub protocol is a decentralized social networking protocol based upon the [ActivityStreams] 2.0 data format. It provides a client to server API for creating, updating and deleting content, as well as a federated server to server API for delivering notifications and content.
(www.w3.org)
Implementers SHOULD include the ActivityPub context in their object definitions.
ActivityStreams:
Implementations producing Activity Streams 2.0 documents SHOULD include a @context property with a value that includes a reference to the normative Activity Streams 2.0 JSON-LD @context definition using the URL " https://www.w3.org/ns/activitystreams".
-
@contextis a recommendation, not a requirement.ActivityPub:
ActivityPub
The ActivityPub protocol is a decentralized social networking protocol based upon the [ActivityStreams] 2.0 data format. It provides a client to server API for creating, updating and deleting content, as well as a federated server to server API for delivering notifications and content.
(www.w3.org)
Implementers SHOULD include the ActivityPub context in their object definitions.
ActivityStreams:
Implementations producing Activity Streams 2.0 documents SHOULD include a @context property with a value that includes a reference to the normative Activity Streams 2.0 JSON-LD @context definition using the URL " https://www.w3.org/ns/activitystreams".
@silverpill aaah, I see. I think we've had this discussion before (or at least I had it with someone else).
For me "SHOULD" falls in the category of the robustness principle: "be conservative in what you do, be liberal in what you accept from others".
So for me if you treat "SHOULD" in a spec as non mandatory you haven't really implemented the spec.
-
@silverpill aaah, I see. I think we've had this discussion before (or at least I had it with someone else).
For me "SHOULD" falls in the category of the robustness principle: "be conservative in what you do, be liberal in what you accept from others".
So for me if you treat "SHOULD" in a spec as non mandatory you haven't really implemented the spec.
@mariusor I don't remember having such discussion. The SHOULD keyword is defined in RFC-2119:
This word, or the adjective "RECOMMENDED", mean that there may exist valid reasons in particular circumstances to ignore a particular item, but the full implications must be understood and carefully weighed before choosing a different course.
There are many valid reasons to not include
@context. We also have almost 10 years of implementation experience and by now full implications are very well understood: by ignoring this recommendation we make messages smaller and developer experience better. No downside at all. -
@mariusor I don't remember having such discussion. The SHOULD keyword is defined in RFC-2119:
This word, or the adjective "RECOMMENDED", mean that there may exist valid reasons in particular circumstances to ignore a particular item, but the full implications must be understood and carefully weighed before choosing a different course.
There are many valid reasons to not include
@context. We also have almost 10 years of implementation experience and by now full implications are very well understood: by ignoring this recommendation we make messages smaller and developer experience better. No downside at all.@silverpill regarding size, ActivityPub is such a verbose protocol that the hundred or so of raw bytes you save through omitting context, are most likely negligible through the prism of connection compression. So to me that's not entirely a "valid reason".
And as developer myself, I think that contexts, even in a non valid JSON-LD implementation, offer enough guidance for building a data vocabulary for them to have plenty of value.
Do you propose we replace contexts with Open API specifications, or how do we coordinate what's a valid vocabulary data object in a federated network? And how do you propose that others discover these specs?
-
@pintoch read this thread?
-
@silverpill regarding size, ActivityPub is such a verbose protocol that the hundred or so of raw bytes you save through omitting context, are most likely negligible through the prism of connection compression. So to me that's not entirely a "valid reason".
And as developer myself, I think that contexts, even in a non valid JSON-LD implementation, offer enough guidance for building a data vocabulary for them to have plenty of value.
Do you propose we replace contexts with Open API specifications, or how do we coordinate what's a valid vocabulary data object in a federated network? And how do you propose that others discover these specs?
@silverpill personally I feel like the various activity/object signing methods that get used in recent FEPs are more egregious from a size point of view, when the in spec behaviour for obtaining canonical versions of a resource is to fetch them from their server, instead of relying on random object signing that introduces so much more friction.
-
@silverpill personally I feel like the various activity/object signing methods that get used in recent FEPs are more egregious from a size point of view, when the in spec behaviour for obtaining canonical versions of a resource is to fetch them from their server, instead of relying on random object signing that introduces so much more friction.
@mariusor@metalhead.club I thought the whole point of signing objects or attaching proofs (none of which I do, mind you) are precisely to save the need to make a new request, which comes with its own overhead.
The good thing is fetching from canonical source will never go out of style.
Aside, it seems like I'm only getting Marius's posts, not silverpills. Makes for an interesting one-sided exchange
-
@mariusor@metalhead.club I thought the whole point of signing objects or attaching proofs (none of which I do, mind you) are precisely to save the need to make a new request, which comes with its own overhead.
The good thing is fetching from canonical source will never go out of style.
Aside, it seems like I'm only getting Marius's posts, not silverpills. Makes for an interesting one-sided exchange
@julian I noticed that your inbox endpoint returns 404s (my activities are delivered to personal inbox, not shared).
-
@hongminhee from the point of view of someone who is "maintaining" a JSON-LD processing fedi software and has implemented their own JSON-LD processing library (which is, to my knowledge, the fastest in it's programming language), JSON-LD is pure overhead. there is nothing it allows for that can't be done with
1. making fields which take multiple values explicit
2. always using namespaces and letting HTTP compression take care of minimizing the transfer
without JSON-LD, fedi software could use zero-ish-copy deserialization for a majority of their objects (when strings aren't escaped) through tools like serde_json and Cow<str>, or System.Text.Json.JsonDocument. JSON-LD processing effectively mandates a JSON node DOM (in the algorithms standardized, you may be able to get rid of it with Clever Programming)
additionally, due to JSON-LD 1.1 features like @type:@json, you can not even fetch contexts in parallel, meaning all JSON-LD code has to be async (in the languages which has the concept), potentially losing out on significant optimizations that can't be done in coroutines due to various reasons (e.g. C# async methods can't have ref structs, Rust async functions usually require thread safety due to tokio's prevalence, even if they're ran in a single-threaded runtime)
this is after context processing introducing network dependency to the deserialization of data, wasting time and data on non-server cases (e.g. activitypub C2S). sure you can cache individual contexts, but then the context can change underneath you, desynchronizing your cached context and, in the worst case, opening you up to security vulnerabilities
json-ld is not my favorite part of this protocol -
@hongminhee from the point of view of someone who is "maintaining" a JSON-LD processing fedi software and has implemented their own JSON-LD processing library (which is, to my knowledge, the fastest in it's programming language), JSON-LD is pure overhead. there is nothing it allows for that can't be done with
1. making fields which take multiple values explicit
2. always using namespaces and letting HTTP compression take care of minimizing the transfer
without JSON-LD, fedi software could use zero-ish-copy deserialization for a majority of their objects (when strings aren't escaped) through tools like serde_json and Cow<str>, or System.Text.Json.JsonDocument. JSON-LD processing effectively mandates a JSON node DOM (in the algorithms standardized, you may be able to get rid of it with Clever Programming)
additionally, due to JSON-LD 1.1 features like @type:@json, you can not even fetch contexts in parallel, meaning all JSON-LD code has to be async (in the languages which has the concept), potentially losing out on significant optimizations that can't be done in coroutines due to various reasons (e.g. C# async methods can't have ref structs, Rust async functions usually require thread safety due to tokio's prevalence, even if they're ran in a single-threaded runtime)
this is after context processing introducing network dependency to the deserialization of data, wasting time and data on non-server cases (e.g. activitypub C2S). sure you can cache individual contexts, but then the context can change underneath you, desynchronizing your cached context and, in the worst case, opening you up to security vulnerabilities
json-ld is not my favorite part of this protocol@hongminhee take this part with a grain of salt because my benchmarks for it are with dotNetRdf which is the slowest C# implementation i know of (hence my replacement library), but JSON-LD is slower than RSA validation, which is one of the pain points around authorized fetch scalability
wetdry.world/@kopper/114678924693500011 -
reposting so @julian sees this
"I noticed that your inbox endpoint returns 404s (my activities are delivered to personal inbox, not shared)." says @silverpill
-
@hongminhee take this part with a grain of salt because my benchmarks for it are with dotNetRdf which is the slowest C# implementation i know of (hence my replacement library), but JSON-LD is slower than RSA validation, which is one of the pain points around authorized fetch scalability
wetdry.world/@kopper/114678924693500011@kopper @hongminhee I'm glad I'm not the only one who noticed this.
-
@hongminhee take this part with a grain of salt because my benchmarks for it are with dotNetRdf which is the slowest C# implementation i know of (hence my replacement library), but JSON-LD is slower than RSA validation, which is one of the pain points around authorized fetch scalability
wetdry.world/@kopper/114678924693500011@hongminhee if i can give one piece of advice to devs who want to process JSON-LD: dont bother compacting. you already know the schema you output (or you're just passing through what the user gives and it doesn't matter to you), serialize directly to the compacted representation, and only run expansion on incoming data
expansion is the cheapest JSON-LD operation (since all other operations depend on it and run it internally anyhow), and this will get you all the compatibility benefits of JSON-LD with little downsides (beyond more annoying deserialization code, as you have to map the expanded representation to your internal structure which will likely be modeled after the compacted one)
