When is initial synchronization done with PROPFIND and when with REPORT sync-collection

[dilyanpalauzov]

DAVx⁵ can do the initial synchronization of a collection in two ways:

• by calling PROPFIND to obtain all elements
• by calling REPORT sync-collection with empty sync-token

Subsequent synchronizations are done with REPORT sync-collection using non-empty sync-token.

The problem is that Cyrus IMAP announces support for sync-collection, but does not implement reports with empty sync-token - cyrusimap/cyrus-imapd#4985.

I have not synchronized a collection for very long time with DAVx⁵, at some moment I wanted to synchronize it. Then DAVx⁵ reported errors, returned by the server, when empty sync-token is used. Afterwards I deleted the account in DAVx⁵ and created it anew.

Apparently DAVx⁵ uses for the initial synchronization sometimes PROPFIND and sometimes REPORT sync-collection. I do not know what is used when. But eventually for CalDAV I convinced somehow DAVx⁵ to use PROPFIND to get all data. However for contacts, I could not achieve the same, so practically my contacts are now not synchronized.

• Describe how to convince DAVx⁵ to do the initial synchronization of contacts with PROPFIND, when the server announces support for REPORT sync-collection.

[rfc2822]

It's described here: https://manual.davx5.com/technical_information.html#synchronization-logic

[dilyanpalauzov]

I proposed a correction for Cyrus IMAP at cyrusimap/cyrus-imapd#4992.

My reading is that under these conditions for CalDAV there is no way to convince it for initial synchronization to use PROPFIND, instead of REPORT with empty sync-token.

[dilyanpalauzov]

I do not understand why this has ever worked. Has DAVx⁵ in the past done the initial request without submitting "<sync-token/>", and later started sending it? In this case, when has DAVx⁵ started sending empty sync-token?

As far as I see in the code, REPORT without sync-token returns/returned all children (as if sync-token was in the request but empty) but the same REPORT fails/failed, if the sync-token is actually present and empty.

[rfc2822]

I can't imagine that this has changed, because the sync algorithm itself (which contains this code) wasn't touched for ages...

[dilyanpalauzov]

This was a libxml2 regression — https://gitlab.gnome.org/GNOME/libxml2/-/issues/783.