On 10/7/25 11:16 AM, Thomas Lamprecht wrote: > Am 07.10.25 um 10:42 schrieb Shannon Sterz: >> one small thought: have you explored whether querying the features of >> the guest when opening the context menu and graying out the snapshot >> option is viable? imo that would be a nicer user experience, but the >> overhead of querying the backend there might be too much. > > FWIW, Nicolas did that in v1/v2, but I wondered w.r.t. slow backend > or spotty and/or high-latency connection making this odd to use. > But I just found that my reply was not CC'ing the list, so I just > re-send it for the record: > https://lore.proxmox.com/pve-devel/[email protected]/ > > That said, given that you suggested the other way around for UX, it > might be indeed better to keep it that way, or at least actually try > how that way works with a slow backend (developer tools can simulate > slow network (the chromium based ones are a bit more powerful in that > regard IIRC), or alternatively use a traffic control (tc) netem qdisc > that adds latency [0] can be added on the PVE server. > Because as mentioned, it could be fine your previous way and the one > Shannon also would prefer, I rather wanted to avoid that we just take > our rather perfect fast and sub-milliseconds lab environments as sole > base for UX decisions. > > [0]: https://manpages.debian.org/trixie/iproute2/tc-netem.8.en.html > e.g. something like: tc qdisc add dev eth0 root netem delay 3000ms
I used some variations of added network throttle using chrome dev tools and traffic control and noticed that the disabling of the button is starting to get really noticable in the 1-1.5 second range (if you don't have sniper level mouse accuracy). Anything below that is negotiable, but if the request stalls even a bit above that range it could become an annoyance (e.g. being just about too early and having to click a second time or simply the long wait). Though similarly, you still have to wait for the window to appear if you clicked it in the new version, and there may also be an error if snapshots aren't supported on the guest. As a note, the original implementation (v1/v2) matches the behaviour found in the Snapshots tab, which has the button disabled until querying snapshot features is complete. This also shows that the `current guest does not support taking new snapshots` until fully loaded. IMO parity between them is essential, as to not have two different approaches which might confuse users. I cannot tell which approach is better here, but it might be a sign that Shannon had also suggested the other way... _______________________________________________ pve-devel mailing list [email protected] https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
