>
> On 08.01.2020 22:51, Stefan Monnier wrote:
> > Yes, I saw that, but it's still not clear to me which part of this costs
> > so much more than tramp-file-local-name itself, nor why.
>
> The whole of it, mostly. I've shown a simple patch which improved
> performance a lot, by Michael says it's unmaintainable (IIUC):


I'm not sure wether what I'll say is relevant to this discussion (please
forgive me if it is not), but I just wanted to mention that a significant
part of "debugging TRAMP problems" in projects like projectile is related
to TRAMP's speed.

I always assumed it was due to the fact that checking file attributes over
SSH is unavoidably slower than doing it locally, but if TRAMP can be
improved for things like `locate-dominating-file` over TRAMP it'd be a huge
help.

Kind regards,
Philippe