Thom Stricklin

That might count as an unintended result, but harmful, really? Bit of an overstatement?

Correct. By default, .gitignore only ignores new files that haven't been added to the repository.

Except xhr.abort() still allows a request to be initiated. Timeout + abort() is probably the most comprehensive solution.

I highly recommend using blowfish/bcrypt versus MD5 or SHA. Unlike the other two, bcrypt is designed to be inefficient, which is what you want for password hashing. While hashing a single password (such as when you set a new password, or compare against a stored value to authenticate) is still relatively fast, with bcrypt, hashing a large number of passwords (such as creating a rainbow table) is going to be many orders of magnitude slower than it would be with MD5 or SHA.

Additionally, bcrypt is tunably complex... You can specify the number of iterations bcrypt should use.

If you're going to be jailing multiple users, you can modify the sshd_config Match block to:

Match group sftp-only

Actually, that's probably how I would do it even with one user, since you're already creating the stfp-only group.

What excalq said... It's a bad idea to rewrite the history of any pushed branch unless you can be absolutely confident that nobody else has pulled or fetched in the meanwhile.

I'm not able to test right now, but a box-shadow with a large spread shoild accomplish the same trick without border-radius issues. YMMV per browser, of course.

It's amazing how many interfaces you can create with hidden checkboxes/radios and adjacent labels. I have found myself using the pattern extensively in all of my recent projects.

I have found that there are some cross-browser issues with this approach. Inputs set to display: none; or zero-size don't seem to select on label-click in older versions of IE--and of course, the :checked pseudo isn't supported. I feel like I was also experiencing problems on IE8? where a ":checked + label" CSS rule was not redrawing immediately.

My workaround has been to absolutely position the inputs (but not the labels) with a high negative left value, and to use JS to explicitly set a class for :checked elements. Workarounds aside, it's a great approach.

@tommy_pyatt: Seems intuitive to me. If this is not intuitive, is there any use of the adjacent selector that is?

Or.... Use min-width for development; convert to min-device-width before your code goes live.

Good advice. I take this a step further and actually define a constant NOW in PHP at the beginning of the request, and in most cases refer to this constant instead of using time().