use cookies to figure out who gave a one click thanks

[GoogleCodeExporter]

[copied from an email thread]

Right now if you send to a group the thank you is anonymous because we
don't know which of them clicked the thank you link.  Since we are
sending one email to a group, we can't give each of them a different
identifiable link to click.  But since clicking the thank you link loads
a page from the feedme site, consider the following hack.

1.  When a mail goes to a single user, put a distinctive url in the
thank-you note (or just remember who it was sent to)
2.  When that thank you note gets clicked, deposit a cookie identifying
the user onto their browser
3.  In the future, check for the identifying cookie when someone clicks
a thank you link.

However, what should we do about email lists, like
[email protected]. What happens if I click thank you from a
haystack email (assigning me a haystack identity), then I click thank you
from an email sent to [email protected]  and another user (so the
browser uses my haystack cookie  identity)? I guess if multiple users who
have already been identified clicks thanks from the haystack email, we
could determine that it's a list, but that relies on accurately identifying
people first. Also, keeping track of email lists may be tricky to implement.

Let's suppose we ignore the problem.  I send a fm to haystack-current
mailing list and someone thanks and gets a cookie associating the
browser with haystack-current (we can think of the cookie as just
containing the unique recipient email address).  so in the future, when
they thank, we can send "haystack-current says thank you"!  Not
terrible.  More complicated is if we _also_ deposited a cookie on that
browser for a different fm addressed to the single user of that
browser.  Now we have two different email addresses cookied with this
browser---how do we know which one is a more "specific" personal
address?    Especially since both or neither might be (one person can
have two email addresses, or they can be on two different mailing lists).

We do have a heuristic: if two browsers clicked thank you on an FM that
went to one address, then it must be a mailing list.  So, we can keep a
blacklist of such mailing lists and ignore cookies from them.  Note that
we _cannot_ use this heuristic if two _different_ FMs to the same
address produce thank you clicks from two different browser, because it
is possible that one user is using and thanking from two different
browsers.  However, I think we can ignore the possibility that one user
is forgetful and says thank you for the same FM from two different browsers.

So: if two browsers thanked the same single-address FM, then the address
is a mailing list.
Otherwise, should we just use the most-recent email-cookie on the
browser?  Or should we use all of them, e.g.
"haystack-current@csail/marcua@csail/[email protected] says thank you!"
Or a blend, say the three most recent email-cookies?

Minor point, but perhaps it is worth taking a mailing-list cookie if we
don't have anything else.  e.g., if the thanker has a cookie for an
address we've identified as a mailing list, and no other cookies, isn't
it better to send use the mailing-list id rather than nothing?  that way
if someone sends to two mailing lists they know which one thanked.

Original issue reported on code.google.com by [email protected] on 27 Apr 2010 at 6:44