foxglove/ws-protocol library has been deprecated and should be updated to foxglove-sdk for eCAL6

[tmorgan4]

Foxglove has deprecated the ws-protocol library and is pushing users to migrate to the foxglove-SDK library which includes a new rust based websocket server. The new websocket server should offer performance advantages and I'm hopeful it solves some of the connection issues experienced with the old ws-protocol websocket server.

I've looked at the ecal-6-pre branch and I'm trying to figure out how to port this over to the new websocket server. I'm stuck since the new websocket server doesn't contain all the same async methods.

Would anyone from the eCAL team be able to help with this? Thanks for all of the great work so far and I'm looking forward to eCAL6!

[KerstinKeller]

Hi @tmorgan4,

yes we are aware of the changes with the foxglove-SDK.
I am currently in the process of further rewrites of the bridge, in order to accomodate a) both eCAL 5 and eCAL 6 versions (using the new eCAL 6 Python API) and b) serving both the ws-protocol as well as the new foxglove-SDK library. (The former is still required for using the foxglove fork lichtblick)

I will keep you posted on the changes.

[tmorgan4]

Thanks @KerstinKeller! Awesome to hear this is already being worked on.

[KerstinKeller]

Hi @tmorgan4,
we have a good start for eCAL6 on the following branch:
https://github.com/eclipse-ecal/ecal-foxglove-bridge/tree/feature/ecal-6-pre
It does not yet support foxglove-sdk, but the server as well as the middleware components have been abstracted, which should make it a lot easier to integrate foxglove-sdk after all.
However, I have not tackled that yet...

[tmorgan4]

@KerstinKeller Great progress on this. Thanks!

I have done some work with foxglove-sdk on protobuf encoding, but I found that the foxglove-sdk python api does not expose a publish_time attribute when logging messages, only log_time. It seems critical to be able to set a message's publish time to the eCAL timestamp (timestamp when message was originally sent via eCAL) but that isn't currently available. I've got a question out to the Foxglove devs and will update if I learn more.

[KerstinKeller]

I previously went the way to try and use eCAL publish time.
The problem which I encountered was that I could not guarantee strictly increasing timestamps over the different channels. E.g. it could be that i receive two different images, "image_front", and "image_back". When forwarding them, you might first forward "image_front" with a timestamp of 10, and then "image_back" with a timestamp of 9, which would cause the UI to reset itself.
E.g. we could only guarantee, due to the distributed characteristics, monotonically increasing stamps per channel, but not overall.

I do not know if Foxglove Studios behavior has changed in that regard, but the frequent resets were totally unacceptable.