Vasilis Papavasileiou

Unfortunately, we do need CHCP (or equivalent). It is an artifact of how Eliom works. If you recompile the application, the service identifiers (and client value identifiers, etc.) may change,...

What @dannywillems suggests could work. It would be convenient for development. But I wouldn't recommend it for the distributable version of the app :).

When you call a pathless service (what used to be called a non-attached coservice), we use a special parameter containing an "identifier" to recognize the service and call the appropriate...

Yes, that would work. The problem with client values and injections remains. We could provide some mechanism for naming them, but there is nothing right now.

Note that there are injections in the code of Eliom itself (and in OS, OT), and these may get in your way... A lightweight OS (template?) is more than welcome....

We have an undeclared dependency on `postgresql-contrib` (to be fixed in `opam-repository`). Please install that.

Why? Users shouldn't really need to deal with `Os_db`, but with the higher-level abstractions (`Os_user` and similar). We don't get anything by intermittently moving code around. Also, "lots of micro-modules"...

Sounds like an anti-spam measure somewhere. Too strict, but beyond our control :). I can't reproduce the 500 with any browser, and my first look into the server logs provides...

That would be a lot of work, including entirely custom code for platforms of marginal importance.

@johnlepikhin, is there a patch somewhere? @balat, while it looks useful generally, do we have a concrete immediate application? Also, would a `React` layer over this be of service?