usememos
feat: add support for server side markdown processing
I have added an option to handle markdown on the backend instead of the frontend. This is for users who prioritize load time and are more limited in computing resources on their devices rather than on the server. For example, if a memos instance is self-hosted and used primarily from a phone. Also, this can be useful for first time accessing public instances using slow internet connection.
This PR is WIP
Also, to avoid dublicating node definition, https://github.com/yourselfhosted/gomark-wasm should be published to the buf repository
@boojack could you please explain the difference between workspace settings and system settings? I'm thinking on where to put server side markdown setting. the same goes for Workspace Profile in workspace_service.proto and Workspace Setting Service in workspace_setting_service.proto.
Thank you for your contribution. But we probably don't need to do this. Here are some reasons or background about our parser.
At the beginning of the design, it was designed as if it was transferring nodes through the backend, but we realized that if there were some large content, the whole API would be very slow to respond and server-side performance would become an issue. So this is definitely not a long term solution, what can be done on the client side should be done on the client side.
We plan to use wasm to put the go parser on the frontend. The results show that it works very well, except that the file size is a bit large. But this problem has been solved (https://github.com/yourselfhosted/gomark-wasm/issues/1)
@boojack I don't quite understand the reasoning on avoiding slow API responses.
my current setup is a private home server with the AMD Ryzen 7 CPU. and I mostly use memos from my phone. obviosly, markdown will be proccessed much faster on the backend, with potential on the future improvments on taking advantage of multithreading and SIMD instructions, which both are not really possible with the browser's WASM runtime. as I see it, using server-side markdown parsing will lead to overall faster rendering. while the API response may take longer, for me as a user, I'm at a net plus in terms of content load time. the negative effect may occur if I host memos on some slow single board PC, such as a RPI, which is slower than most of my client devices. but that's why I would like to introduce this feature as an option
despite all this, after reducing the size to just 1.1mb, this PR might be not as relevant, since most of the problems with the WASM blob were because of it's size. so I understand if you prefer to not spend additional resources supporting this feature in the future
Implemented with https://github.com/usememos/memos/commit/155c5baf2c7da865f75c337eaacf64fd9170a9f9