Rework host/hostname/authority implementation. #1561
Conversation
|
This is still a draft/prototype. |
ioquatix
force-pushed
the
host-authority
branch
from
a618e48
to
6bc2a70
Compare
ioquatix
force-pushed
the
host-authority
branch
5 times, most recently
from
d710df5
to
efede61
Compare
|
This is quite a complex change, so I welcome a little bit of scrutiny. That being said, the vast majority of the specs are passing without changes, so that's a good sign. |
…ame`.
u = URI("http://[::1]/bar")
p u.hostname #=> "::1"
p u.host #=> "[::1]"
With IPv6, any time we have a string which represents an authority, as defined by RFC7540, the address must be contained within square brackets, e.g.: "[2020::1985]:443". Representations from the `host` header and `authority` pseudo-header must conform to this format. Some headers, notably `x-forwarded-for` and `x-forwarded-host` do not format the authority correctly. So we introduce a private method `wrap_ipv6` which uses a heuristic to detect these situations and fix the formatting. Additionally, we introduce some new assertions in `Rack::Lint` to ensure SREVER_NAME and HTTP_HOST match the formatting requirements.
ioquatix
force-pushed
the
host-authority
branch
from
efede61
to
aa68f07
Compare
| def hostname | ||
| host, address, port = split_authority(self.authority) | ||
|
|
||
| return address | ||
| end |
There was a problem hiding this comment.
This looks like inconsistency: the method hostname returns address… why? I think we should name it either hostname or address ecerywhere.
There was a problem hiding this comment.
It matches the implementation of URI. I'm not saying it's right. I just decided to follow the existing design/method names. But you are right. Maybe it should be address or something else...
There was a problem hiding this comment.
I mean, if you even just rename local variable to hostname — it'd be better. I don't see address as URI method or in its sources.
There was a problem hiding this comment.
Understood. It is actually confusing to me too in a way.
u = URI("http://[::1]/bar")
p u.hostname #=> "::1"
p u.host #=> "[::1]"
What does make sense to me is when you say "host_with_port". That means, the "escaped" host and :port appended. You can't have hostname_with_port. It's simply not possible because it becomes ambiguous to parse.
So, there's that logic... But yeah, I also see it's confusing.
There was a problem hiding this comment.
Yes, we could rename it hostname everywhere. I see what you are saying, and I don't think it would be a problem. Except that technically, I don't even know if 192.168.1.1 is a hostname? Or 1:: is a hostname? To me, something like www.google.com is a hostname. Not sure what is the right word/definition here.
There was a problem hiding this comment.
I know it's about local variable name but the entire thing is kind of confusing, so we should try to get it right for our users or we just perpetuate the chaos.
There was a problem hiding this comment.
So we have several options:
1/ Rename local variable address -> hostname and return address to return hostname.
2/ Rename method def hostname to def address.
3/ Remove def hostname.
4/ Add def uri which returns an instance of URI which represents what the URI was used to make the request.
There was a problem hiding this comment.
1 is like compliance with URI from stdlib.
2 is like try to improve naming, but, as I see, still with logic violations.
3 is radical, and, I guess, we still need for a helper method. If not — the best code is no code.
4 seems interesting, and "URI" sounds common enough for IPs and domains, especially with URI instance inside.
So, I can't say before trying, but I'd try to make in this order: 3, 4, 1, 2.
There was a problem hiding this comment.
I'll wait for feedback from @jeremyevans before I make any further changes.
There was a problem hiding this comment.
Just for the record, there are several new methods, and that includes def hostname. It's not released in Rack 2.1.
|
Thanks @jeremyevans I will update the specs according to your feedback. |
ioquatix
force-pushed
the
host-authority
branch
from
63119ed
to
40fb1f4
Compare
ioquatix
force-pushed
the
host-authority
branch
from
40fb1f4
to
eec2635
Compare
Attempt to rework the host/hostname/authority methods to better follow the RFCs regarding host/authority, along with URI specifications for host/hostname. Fixes #1560.