Re: [ITP] FUSE 2.8

[Bill Zissimopoulos <billziss@navimatics.com>]

On 7/22/16, 4:59 AM, Adrien JUND wrote:


>The package should be renamed winfsp-fuse for give ability of cygwin
>users to choose which solution they would like to use. Like
>dokan-fuse, cbfs-fuse and other projects that offer the same
>service...

I am not opposed to renaming the package if that’s what the Cygwin
community wants.

However I note that this may create the following problem. Packages that
depend on FUSE will need to have that dependency satisfied by a number of
different *-fuse packages. Perhaps the dependency system in Cygwin is
flexible enough to support this (I don’t know).

For example I am planning to follow up this submission with SSHFS and
FUSEPY submissions. I would like to make these packages depend on a FUSE
package, so that we do not end up with a winfsp-sshfs package, a
dokan-sshfs package, etc. Even more importantly I would like SSHFS and
FUSEPY to build/run out of the box without having to add special logic for
winfsp-fuse vs dokan-fuse, etc.


>The official fuse window integration is an official request made by
>devs on WPDEV. This request is well placed on the top so it is
>probably only a question of time before windows do it in the same time
>as the Linux subsystem integration.
>https://wpdev.uservoice.com/forums/266908-command-prompt-console-bash-on-u
>buntu-on-windo/suggestions/13522845-add-fuse-filesystem-in-userspace-suppo
>rt-in-wsl

I am sorry, but having worked for Microsoft this is definitely wishful
thinking. If anything Microsoft would likely do it themselves. Heck I
might just rejoin them and do it for them ;)

>I also would like to point out that WinFSP has absolutely no feedback
>of any kind by users and has not been tested on all windows versions.
>I think Kernel drivers should at least have some feedback and known as
>used in production before choosing to be distributed as cygwin
>package.
>Unstable kernel drivers can create severity damage in case of BSOD
>like windows or user files corruption.
>
>These analyses are probably severe but for the good of cygwin users,
>integrate kernel driver dependence should be well thought before
>making the step.

Rubbish!

I am sorry that you are so upset Liryna that I decided to create my own
project instead of continuing to contribute to Dokany. But to go around
and spread unsubstantiated rumors is not acceptable.

You do not know how many people use WinFsp. To claim that there is no
feedback is simply false. There is no feedback that you see, but there is
quite some favorable feedback that I have seen including commercial
propositions. Furthermore being the number 1 trending project on GitHub
(for the C language) certainly shows some favorable feedback.

I would strongly argue that WinFsp is stabler and takes file system
semantic correctness far more seriously than Dokany. I note that you use
my own file system test suites to test your file systems and that Dokany
breaks with such basic things like memory mapped files and share access. I
could enumerate all of Dokany’s failings, but I will spare you the
embarrassment.

Bill