Exclusions not recursively applied

[mumen]

I can't understand the reason why no mirror solution are true mirrors solutions when dealing with exclusions.

Let's look at the problem with an example:

Imagine that I prepare of the parameters for a mirror copy of my datas that I daily use. Later I remark that some files should be excluded to the destination. So I add a new corresponding rule to the filters and reapplies the synchronisation.

You may know what is happening then: the supposedly excluded files will forever remain in the destination folder. I will have to delete them manually in order to have an exact mirror.

In those conditions, one can never be totally sure that the destination exactly corresponds to the source with the rules adopted, what is confusing and can conduct to over consumption of destination disk space.

No sync soft I tried, even RoboCopy, had a correct comportment for this situation.

I can't manage to find a reason why such a choice should be imposed, nor technical dead-end, nor usefulness behaviour. On the contrary, one should explain to me why it is better to keep things henceforth unwanted in the destination. If only one software was making that mistake, I'll understand, but all of them? I'm stunned!

I think the synchronisation should always be done AFTER applying filters, not before. Parametrize this choice is not even useful! For now this is the first problem to me. I know I'm purist and that this is a secondary problem. But to me, the sync process must absolutely be clear to the user, in the least details, or else the trust is in danger.

Another reason, indirect, for correcting this problem, is the excellent way FFS is making the version management. This is so interresting that it seems possible to universalise the mirror way, in replacing the usual incremental backup.

Let's be clear: from what I've seen, FFS seems at the top for the synchronisation process clarity (even if I still have to deeply test the "two ways" method). From my point of view leaving such a (even little) source of confusion is damaging the process of creating and maintaining an efficient sync/backup strategy by mirroring folders.

[Zenju]

There are two main reasons for this design:

1. File exclusions are happening during comparison and they affect both left and right hand side equally. What FFS can't see, will not be touched later during sync.

2. If file exclusions were not symmetrical, this could be risky or surprising for sync variants other than mirror. For example FFS could try to copy a file that it sees on one side to the other, which is empty (but in reality the file is really there but was just excluded) and fail with "target already existing" although the FFS GUI showed otherwise.

With these considerations, what are the solutions/alternate designs?

[Zenju]

Related feature request:
[404, Invalid URL: https://sourceforge.net/p/freefilesync/feature-requests/279/]