It's unfortunate that we have this mess. But couldn't this have been avoided by defaulting to minimal access? Per Mozilla (https://developer.mozilla.org/en-US/docs/Web/HTTP/Cookies), if a cookie's domain isn't set, it defaults to the domain of the site excluding subdomains. If instead, this defaulted to the full domain, wouldn't that resolve the issue? The harm isn't in allowing people to create cookies that span sites, but in doing so accidentally, correct? The only concern is then tracking cookies. For this, a list of TLDs which it would be invalid to specify as the domain would cover most cases. Situations like github.io are rare enough that there could simply be some additional DNS property they set which makes it invalid to have a cookie at that domain level.
Similarly, the secure and http-only properties ought to default to true.
Since first-party cookies are used to track login state, you do want someway to opt into sharing them site wide for sites where forum.example.com and mail.example.com have the same concept of a user.
For this, a list of TLDs which it would be invalid to specify as the domain would cover most cases.
That is what the public suffix list is?
some additional DNS property they set which makes it invalid to have a cookie at that domain level
The problem with his approach is that registries would need to opt in, but in some cases registries can't be bothered and, since it's a security matter, the PSL maintainers (Mozilla) do it for them.
Are
forums.example.com and
mail.example.com
the same site? I'd say yes, since they're probably run by the same
people. What about
example-a.github.io and
example-b.github.io? I'd say no, since GitHub
allows anyone to register pages
like
username.github.io. I can make my judgments
as a human, but what should the browser do? Should
www.example.com be able to set a cookie that will be sent
to
mail.example.com?
It is a bit of a hack, but the way browsers deal with this is a big
list: the Public Suffix List.
The PSL contains, for example, com and
github.io, which tell us that example.com
and example.github.io are independent sites. On the
other hand, any subdomains are not separate sites:
forums.example.com and mail.example.com.
Have a look, it's pretty hairy: public_suffix_list.dat
Browsers are somewhat ashamed of the hackiness of site,
and nervous about the security risk of omissions, and so have
generally used a much stricter concept of origin when
introducing functionality. For example,
https://a.example.com cannot write to
localStorage in a way visible to
https://b.example.com. As browsers work to prevent
cross-site tracking, however, with privacy changes such as cache partitioning, the
origin model is too strict. These mitigations generally
use the PSL, and I wanted to look back at its origins.
HTTP was originally completely stateless. This poses challenges if
you want to implement per-user functionality, like a shopping cart.
Netscape's solution, which the world adopted, was cookies. If you
read the
original
specification, it has some discussion of how to prevent someone
setting a cookie on all of .com:
Only hosts within the specified domain can set a cookie for a domain and domains must have at least two (2) or three (3) periods in them to prevent domains of the form: ".com", ".edu", and "va.us". Any domain that fails within one of the seven special top level domains listed below only require two periods. Any other domain requires at least three. The seven special top level domains are: "COM", "EDU", "NET", "ORG", "GOV", "MIL", and "INT".This simple heuristic worked reasonably well at the time: it understands that
example.com and
example.co.uk are independent sites, separate from other
.com or
.co.uk sites.
Perhaps because this special-cased domain names, it was not included in the first two attempts to standardize cookies, RFC 2109 (Feb 1997) and RFC 2965 (Oct 2000).
There were, even from the beginning, cases that this heuristic did not
handle. My library growing up was mln.lib.ma.us, which
ideally would not have shared cookies with anything else under
lib.ma.us. In 2000, however, ICAAN announced
seven more TLDs, and initially browsers did not allow anyone to set
cookies on example.info etc. It wasn't too bad, since
you could still set a cookie on www.example.info, but you
couldn't share it with forum.example.info.
In 2005-2006, Mozilla decided to replace their inconsistent collection
of heuristics and exceptions with an explicit list (b319643,
b331510),
effective_tld_names.dat. You can see the first public
version on
github (Mar 2007).
The next round of cookie standardization, RFC 6265 in 2011, recommended projects use it:
NOTE: A "public suffix" is a domain that is controlled by a public registry, such as "com", "co.uk", and "pvt.k12.wy.us". This step is essential for preventing attacker.com from disrupting the integrity of example.com by setting a cookie with a Domain attribute of "com". Unfortunately, the set of public suffixes (also known as "registry controlled domains") changes over time. If feasible, user agents SHOULD use an up-to-date public suffix list, such as the one maintained by the Mozilla project at <http://publicsuffix.org/>.
This still doesn't explain how github.io got on the list:
that's not a public registry, the way co.uk is. The
first private registry to be added was operaunite.com, in
November 2009 (b531252):
The domain operaunite.com is used by Opera's new Unite feature (a small web server built into Opera 10.10,http://unite.opera.com/). Each instance of Opera Unite have a nameserver.username.operaunite.com. While some restrictions are being implemented in Unite, there are still some ways to set cookies for theoperaunite.comdomain, and we would like to restrict the impact by adding this domain to the public suffix list.
Next were appspot.com for App
Engine and blogspot.com for Blogger (b593818),
though the Blogger change was rolled back for two years (b598911,
b805367).
These changes seem to have been uncontroversial; I don't see any
pushback about how these are not "real TLDs".
As more of these came in, there was discussion about how these
were fundamentally different concepts (b712640,
2011), and the list was split into public ("BEGIN ICANN DOMAINS") and
private ("BEGIN PRIVATE DOMAINS") sections. For example, no one
should be able to get a wildcard cert for *.co.uk, but
one for *.github.io still makes sense.
Over the last ten years, I believe everyone has migrated to using Mozilla's list. It does take some time for updates to fully propagate, since the list is compiled into browsers, but having one place to update and one place to check for the definition of a site is pretty good.
Comment via: facebook