04:16 <kentonv> isd: Hmm. I guess the security properties are no different from WebSocket. So I suppose it would be reasonable.

04:16 <kentonv> isd: However, you may find that if your Sandstorm server is behind any kind of proxy (forward or reverse), it ruins everything...

04:25 <isd> I will open an issue to track it. The proxy stuff could cause hiccups, but the current solution isn't terribly pleasant. I'd expect to keep supporting the websocket approach, since I don't know that every IRC client out there can be easily coaxed into talking to an http proxy (esp. on a per-network basis).

04:31 <isd> https://github.com/sandstorm-io/sandstorm/issues/2902

21:04 <TMM> hi all! Another capnp question: Is it possible to distinguish between a default value and an unset value?

21:16 <isd> TMM: two questions: (1) why? (2) I assume you're talking about the C++ implementation?

21:18 <isd> (I'm always thrown by (2); I mostly use the Go implementation and am hacking on a haskell one, so my first instinct is to answer in terms of the data format itself, forgetting that people are using a library. Very different questions)

21:18 <TMM> isd, well, 'why' is I have a message for updating one of a set of fields in a struct, say you have a user and you want to update only its first_name field. It'd be nice if I can tell what fields to update

21:18 <TMM> isd, and yes, c++

21:19 <isd> is first_name Text?

21:19 <isd> Pretty sure there's a HasFoo() or such if the type is a pointer type (like Text)

21:20 <isd> If it's not a pointer type, you're talking small amounts of data anyway, so my first instinct is just set it unconditionally until you've proven it's a problem.

21:20 <TMM> isd, data size I don't care about so much, it's more detecting whether or not it's set

21:21 <TMM> but if Text is a pointer and I can just do HasFirst_Name() that would help

21:22 <isd> For non-pointer types, whether it's "set" or not is just a question of whether there's space allocated in the segment for it. I don't know what it does if you try to set it and there isn't.

21:25 <isd> logic for getters is something like if(sizeof dataSection >= offsetof myfield + sizeof myfield) return readValue(); else return defaultValue();

21:25 <isd> I don't know if the C++ library exposes what you need to check that or not. I also still don't understand what you're trying to do.

21:26 <isd> There was some recent-ish discussion on the mailing list about this...

21:27 <TMM> Well, what I'm trying to do is send a 'User' struct to my endpoint with a message that this user should be updated

21:27 <TMM> but I need to know which of the fields I actually need to update

21:30 <isd> Here's the thread: https://groups.google.com/forum/#!topic/capnproto/TLQzbJ2oCXE

21:31 <isd> Short answer is, for non-pointer types, you can't do that.

21:31 <isd> Also some suggestions for alternative approaches.

21:31 <isd> If it's a pointer type, you can check with Has* as I mentioned.

21:31 <TMM> awesome, thanks

22:00 <isd> no problem

22:23 <TMM> isd, would you mind having a look at this? http://paste.debian.net/924851/

22:23 <TMM> I get the following output for that program: terminate called after throwing an instance of 'kj::ExceptionImpl'

22:23 <TMM> what(): src/capnp/serialize.c++:43: failed: expected array.size() >= offset; Message ends prematurely in segment table.

22:23 <TMM> I have no idea what's wrong with that code

22:26 <pdurbin> isd: I'm looking forward to seeing your lightning talk on Sandstorm at https://media.libreplanet.org/u/libreplanet/

22:28 <TMM> isd, nvm, ArrayPtr wasn't what I thought it was

22:49 <TMM> Now I'm trying to figure out how to *set* a Void pointer in a message with setField() where Field is of type Void in the message format

23:05 <TMM> isd, if you have time, could you point me to the documentation on how to get a capnp::Void from a messagebuilder?

23:16 <TMM> oh, I had Void, not AnyPointer

23:16 <TMM> that's probably why

23:40 <TMM> Figured that out too :)

23:40 <TMM> ladieda, sorry for the noise