Did anything every come of this?

On Sunday, June 16, 2019 at 6:45:40 AM UTC-7, Mark S. Miller wrote:
>
> (From Martin BZ and MarkM jointly)
>
> Hi Kenton, could you look at 
>
> https://github.com/wanderer/WASI/tree/master/schema
>
> We want to suggest this to the WASI folks, to suggest that this be the way 
> WASI IDLs should be specified. Their current plans are to start with 
> WebIDL, which is rather horrible. So we'd like to send this to them as soon 
> as reasonable, but not sooner. So, at this stage, we're not looking for 
> detailed feedback on particulars, but rather, whether this is a reasonable 
> style for using the Cap'n Proto schema language.
>
> Thanks!
>
>
>
> On Sun, Jun 16, 2019 at 10:37 AM Mark S. Miller <[email protected] 
> <javascript:>> wrote:
>
>> [+Michael, +Dan] 
>>
>> I like this direction!
>>
>> Where can I find a bnf for the Cap'n Proto schema/idl language? I want to 
>> start from there to create a concrete syntax for the Chainmail (formal 
>> specification language for ocap patterns) abstract syntax. I had previously 
>> started with a Jessie-like syntax, but I think Cap'n Proto would be a 
>> better place to start. As part of this, I want to write specifications for 
>> the ERTP and SwingSet APIs as Cap'n Proto files that I can imagine would 
>> grow into Chainmail specs. Together with WASM/WASI bindings for Cap'N 
>> Proto, this would be a start for enabling non-JS WASM instances to plug 
>> into our SwingSet kernel.
>>
>> To close this loop, we should define the subset of Cap'n Proto that we 
>> expect to compile to WASM/WASI bindings. SwingSet should then restrict 
>> itself to that subset.
>>
>> Is there a standard representation of the AST that a Cap'n Proto 
>> schema/idl file parses into? With such an AST translated to JSON, I would 
>> also like to explore changing our marshaling system so that it only allows 
>> through calls described in the IDL and made available to the comm system at 
>> runtime via a derived JSON representation suitable for that purpose. That 
>> would put us in a good position to explore eventually cerealizing to Cap'n 
>> Proto itself.
>>
>>
>>
>> On Sat, Jun 15, 2019 at 6:46 PM 'Kenton Varda' via Cap'n Proto <
>> [email protected] <javascript:>> wrote:
>>
>>> Hi mjbecze,
>>>
>>> This is interesting! Is this related to:
>>>
>>> https://github.com/WebAssembly/design/issues/1274
>>>
>>> ?
>>>
>>> On Sat, Jun 15, 2019 at 9:08 AM <[email protected] <javascript:>> wrote:
>>>
>>>> Hello, 
>>>> I'm currently investigating whether capt'n proto schema would be 
>>>> appropriate for describing Webassebly interfaces. I'm attempting to 
>>>> translate this interface 
>>>> <https://github.com/WebAssembly/WASI/blob/master/docs/WASI-overview.md> 
>>>> to capn proto schema. I several questions and I'm sure I'll have more as 
>>>> the work progresses.
>>>>
>>>> 1. Webassebly will soon have opaque references 
>>>> <https://github.com/WebAssembly/reference-types>. How would you 
>>>> represent reference types in capt'n proto?
>>>>
>>>
>>> Cap'n Proto's "Capability" type is exactly what you want here: it 
>>> represents an external reference to "something". Note that interface types 
>>> in Cap'n Proto are all subclasses of "Capability" -- i.e. "Capability" is 
>>> kind of like Java's "Object". But it's totally valid to use just the type 
>>> "Capability" to represent a handle to some opaque thing that has no 
>>> particular methods you can call.
>>>
>>> Also worth noting that in Cap'n Proto's real RPC system, capabilities 
>>> are referenced through table indices, just like Wasm opaque references 
>>> apparently will be. Check out:
>>>
>>>
>>> https://github.com/capnproto/capnproto/blob/77f20b4652e51b5a7ebda414e979e059a6c7c27c/c++/src/capnp/rpc.capnp#L116
>>>
>>> Also, Cap'n Proto's C++ implementation has the ability to attach file 
>>> descriptors to capabilities and pass them between processes (via SCM_RIGHTS 
>>> on unix sockets):
>>>
>>> https://github.com/capnproto/capnproto/pull/821
>>>
>>> So there's lots of precedent here.
>>>
>>> 2. Is annotation the correct way to represent pointers? for example
>>>>
>>>
>>> Oof, this is pretty awkward, especially in that it presumes the callee 
>>> has the ability to reach into the caller's address space. Obviously, Cap'n 
>>> Proto hasn't been designed with that sort of thing in mind.
>>>
>>> With that said, I like your second solution, transforming return values 
>>> into things that are stored to the pointer. This allows the schema to 
>>> express the semantics in a way that makes sense for classic RPC, and then 
>>> return-by-write-to-pointer becomes part of the "calling convention" that 
>>> maps Cap'n Proto to Wasm.
>>>
>>> -Kenton
>>>
>>> -- 
>>> You received this message because you are subscribed to the Google 
>>> Groups "Cap'n Proto" group.
>>> To unsubscribe from this group and stop receiving emails from it, send 
>>> an email to [email protected] <javascript:>.
>>> Visit this group at https://groups.google.com/group/capnproto.
>>> To view this discussion on the web visit 
>>> https://groups.google.com/d/msgid/capnproto/CAJouXQ%3DYWW-L4Rp5JZHD5CGTbhiKypWuoP5HJ10CxKqL8A6Z%3Dg%40mail.gmail.com
>>>  
>>> <https://groups.google.com/d/msgid/capnproto/CAJouXQ%3DYWW-L4Rp5JZHD5CGTbhiKypWuoP5HJ10CxKqL8A6Z%3Dg%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>> .
>>>
>>
>>
>> -- 
>>   Cheers,
>>   --MarkM
>>
>
>
> -- 
>   Cheers,
>   --MarkM
>

-- 
You received this message because you are subscribed to the Google Groups 
"Cap'n Proto" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/capnproto/e9ee73e2-9950-45f6-ba83-c8864f217a32%40googlegroups.com.

Reply via email to