Even though I agree that this is a desirable feature for Preview, it
would have been nice to see it on the PReview plan so we could have
scheduled time for it.
On Mar 22, 2007, at 7:37 PM, Katie Capps Parlante wrote:
Hi Heikki,
I agree that a password encryption feature is something that we're
going to want, and given that you have it working on your branch it
makes sense to incorporate it. Apologies for not speaking up earlier.
I will acknowledge (for PPD, QA teams) that this could be
considered minor feature creep, adding QA load, etc. We need to
make sure we have a spec, in particular for testing. Can you help
with that?
You ok with this Sheila and Aparna?
We do need a spec to test this out. So, yes as long as we get one we
should be fine.
More comments in line...
Heikki Toivonen wrote:
Chandler has the ability to remember passwords, and many high profile
programs (e.g. Firefox) that have this ability can encrypt these
passwords.
Doing encryption/decryption like this traditionally requires the
user to
set a master password. The master password is never stored on
disk, it
will be asked from the user on demand, and may be remembered in
memory
until program shutdown or timeout.
I think we need to provide some level of encryption support in
Preview
timeframe. For example, I think our users should be able to submit
their
repositories to us for debugging purposes without us learning their
passwords.
You make a good point here. Mimi pointed out offline that we still
might be stuck if a user only decides they want to send us their
repository after they're in a situation where Chandler has wedged
-- at this point it would be too late to set up the master
password. Nonetheless, it will prove useful in some situations.
Do we want to default to requiring a master password to encrypt and
decrypt the other passwords?
I think you have it right in your implementation: by default, no,
not required.
+1
Or do we start unencrypted, offer a "encrypt" checkbox in the
accounts
dialog, and also when making a repository backup/dump? (I think I am
slightly in favor of this.)
Yup, perhaps we could hear from Mimi the best place for this.
Do we want to provide encrypting arbitrary items/attributes? (I
wouldn't
worry about this until after Preview.)
Yup, sounds like feature creep right now. Perhaps best to wait
until we hear some user asking for the feature. :) Any thoughts Mimi?
Do we want to protect the passwords in memory? I must point out that
this would be quite a bit of work, and it is not certain we could
even
cover all cases (passing password strings into libraries we may
not have
control over, for example). This would involve things like: clear out
master password on timeout, never store the other passwords in clear
text except for the moment when they are needed, zero out the actual
bits in memory once done, prevent password memory from being swapped
out, etc. (I wouldn't worry about passwords in memory myself.)
Agree with your assessment.
Please note that Chandler already supports encrypting the entire
repository. An alternative on some operating systems is to ask the
OS to
encrypt the disk/directory where the repository is.
Yes, perhaps encrypted filesystems are the long term solution for
people who really care about this.
Another thing to note is that many OSes provide password safes of
their
own with naturally platform specific APIs. I am not suggesting we
try to
hook up with these in Preview timeframe.
Yup, I agree with other conversations in the thread that this is
the direction we want to go in, and too expensive for Preview.
Perhaps this would be a good "helpus" project?
Cheers,
Katie
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "Design" mailing list
http://lists.osafoundation.org/mailman/listinfo/design
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "Design" mailing list
http://lists.osafoundation.org/mailman/listinfo/design
