| 11 Jun 2019 |
 Maximus | or is 403 not a valid status code? trying to make sense of the whole thing | 19:02:55 |
 JulianF | ]y | 19:02:53 |
| Maximus | Jason: shush about non-identity stuff! | 19:03:06 |
 Jason | In reply to @travis:t2l.io
it sounds like Synapse is being correct and not accepting historical characters for new identifiers, which is correct. That's ... not what I'm asking or talking about... | 19:03:09 |
| Maximus | get a room (ah!) with Travis :p | 19:03:21 |
| Jason | @max:kamax.io: It's not possible, we're not allowed to communicate by papal decree | 19:04:01 |
| Maximus | then wait until I'm done sorting out the unbind stuff | 19:04:22 |
| Jason | sure np | 19:04:29 |
| Jason | I'll wait until you're done getting fucked over by NV | 19:04:42 |
| * Jason rubs hands together | 19:04:49 |
 TravisR | Maximus: 403 is valid, and tbh we should just put that in the spec. The error should be proxied to the client, however it is not mandatory. Therefore mxisd can return 403 and Synapse can return 502 and both are still compliant, so long as they return standard Matrix errors. Ideally though, the spec says 403 is valid and Synapse proxies the error - I've opened https://github.com/matrix-org/synapse/issues/5426 and https://github.com/matrix-org/matrix-doc/issues/2117 to cover this, hopefully. | 19:05:42 |
| Maximus | So, given synapse code:
- Nothing will ever be shown to the client, so the user has no mean to know what is going on (was always the case and the biggest issue with this situation)
- 400, 404 and 501 log a warning, anything else throws a 502 which actually turn into a 500 with a generic error (so nothing is actually kept)
| 19:06:06 |
| Maximus | so right now, nothing is working any better than before, and it seems like you're confirming with those two news issues that there is no workaround either | 19:06:42 |
| Maximus | so there is still nothing mxisd can do about this | 19:07:05 |
| TravisR | if that's how you want to interpret what I'm saying, I can't change your mind. | 19:07:22 |
| Maximus | I'm trying to find a way to make this work | 19:07:45 |
| TravisR | if mxisd's goal is to prevent usage of /unbind, it should return one of the 3 errors which allow it to do so safely. | 19:08:06 |
| TravisR | if it only wants to prevent some usages of /unbind, it should return 403 | 19:08:27 |
| Maximus | In reply to @travis:t2l.io
if it only wants to prevent some usages of /unbind, it should return 403 This is the use case (and always has been) wants to support | 19:08:59 |
| TravisR | 403 it is then | 19:09:25 |
| Maximus | so can you confirm, given the source code, there is no means currently for mxisd to have a message showns to the user? given that 400, 404 and 501 are only logged, and 403 would throw a generic error | 19:09:45 |
| TravisR | you're asking me to confirm something that you can figure out on your own | 19:10:17 |
| TravisR | sorry, I'm not doing that for you. | 19:10:22 |
| Maximus | Alright | 19:10:43 |
| Jason | @max:kamax.io: On topic, let me know what you want construct to implement to convey errors to the user for you | 19:11:06 |
| Maximus | just relaying the error message from the standard matrix error object is all that is needed, regardless of the status code | 19:11:42 |
| Jason | oh. construct just does that naturally... | 19:12:21 |
| Jason | Usually it just appends to the errmsg string as the stack unwinds, so mxisd's string would probably be at the end of the final message | 19:13:18 |
| Maximus | sure, as long as it's visible, even if something else is added, is totally fine | 19:15:37 |
| Jason | In reply to @travis:t2l.io
sorry, I'm not doing that for you. Is this a true technical issue or is this just a case of giving people who matthew doesn't like a hard time to curry favor with the boss because the culture at NV values such behavior | 19:16:24 |
