WebClient service
Everything NcDavTray maps goes through the WebClient service — the Windows WebDAV mini-redirector. It is a part of Windows, it is old, and its defaults are conservative in ways that show up as odd failures. The WebClient tuning tab makes its settings visible and changeable without a registry editor.
Tip
TL;DR — Settings → WebClient tuning. Reading the values needs no admin rights; only Apply changes asks for elevation. The two settings that most often need raising are the maximum file size (default is far below 4 GB) and the files per folder limit.
The WebDAV redirector turns a WebDAV server into something Explorer treats as a network drive. It is what makes Z: behave like a drive rather than like a web page — and it is also where the limits live that people run into: file size caps, timeouts, and how many entries it will hold for one folder.
It has to be running, and its startup type must be Manual or Automatic. Manual is enough — Windows starts it on the first WebDAV access.
A disabled WebClient service cannot be started on demand, so no WebDAV drive works at all. The tuning tab can change the startup type and start the service for you, which needs administrator rights.
Note
On some systems the service is disabled by policy rather than by setting. Then the change will not stick, and the decision belongs to whoever administers the machine.
| Setting | Why it matters |
|---|---|
| Maximum files per folder | the attribute cache limit; a folder with more entries than this behaves erratically in Explorer |
| Maximum file size | the redirector refuses larger transfers outright. Can be raised to 4 GB |
| Local server timeout | how long to wait for a server Windows counts as local. Which almost certainly means never — see below |
| Internet server timeout | the same for everything else, which in practice is every server you will ever mount |
| Send / receive timeout | how long a single transfer may stall before it is given up |
| "Server not found" cache lifetime | how long Windows remembers that a server did not answer — this is why a server that came back can still appear dead for a while |
| Service startup type | Disabled, Manual or Automatic |
| Start / restart the service | without going to the Services console |
Every value is shown with what is currently in the registry and a readable explanation of what it does, plus an inline help text.
Tip
If a server that is demonstrably back still shows as offline for a minute or two, the server not found cache is usually the reason. Lowering it makes recovery after an outage feel much quicker.
Two of the timeouts do the same job for two kinds of server, and the redirector picks one of them per connection. Microsoft documents both values but never says how that pick is made, and the answers you find elsewhere borrow a rule from a different Microsoft article about credential prompting. So this was measured instead of copied.
The host name decides, and nothing else.
| The host in the UNC path | counts as | uses |
|---|---|---|
no dot in the name, such as nextcloud
|
local server | Local server timeout, default 15 s |
a dot anywhere, such as cloud.example.com
|
internet server | Internet server timeout, default 30 s |
an IP address, such as 192.168.1.10
|
internet server | Internet server timeout |
An IP address counts as an internet server because it contains dots. That is the whole rule.
Security zones make no difference. Adding a server to Local intranet under Internet Options changes nothing about which timeout applies. That was checked in both directions: a dotted name forced into the intranet zone still used the internet timeout, and a dotless name forced into the internet zone still used the local one.
How this was measured
Two host names, one with a dot and one without, were pointed at the same unused address on the local network, so that every connection attempt runs into its timeout instead of reaching anything. The two timeouts were set far apart, at 5 and 12 seconds, both below the point where TCP gives up on its own. The dotless name failed after 8 s, the dotted one after 16 s, reproducibly, with only the name differing. Swapping the two registry values swapped the two measured times exactly, which is what ties the behaviour to these settings rather than to something else.
Note
NcDavTray connects over HTTPS only, and no public certificate authority may issue a certificate for a single-label name — those are barred as internal names. So unless your server has a certificate from a CA of your own, every mount you make uses the internet timeout. This is why the Local server timeout field stays greyed out until one of your mounts actually names a host without a dot.
Reading the current values requires nothing. Apply changes writes to HKLM and therefore asks for elevation — the button carries the UAC shield to make that visible before you press it.
Caution
These settings are global to Windows and affect every WebDAV drive on the machine, not only NcDavTray's. That is a property of the redirector: there is one set of values, shared by everything that uses it.
