Hackney support for float status codes

Questions / Help
1

Is there any solution / workaround for hackney when status codes are returned as a float? ex: 401.1

when I try to run HTTPoison.get("http://www.lowiqcanadian.com/id2.html") inside of iex i get

** (ArgumentError) argument error
    :erlang.list_to_integer('401.1')
    (hackney 1.17.4) /Users/k/v-repo/fetchers/deps/hackney/src/hackney_http.erl:221: :hackney_http.parse_reason/4
    (hackney 1.17.4) /Users/k/v-repo/fetchers/deps/hackney/src/hackney_response.erl:74: :hackney_response.wait_status/1
    (hackney 1.17.4) /Users/k/v-repo/fetchers/deps/hackney/src/hackney.erl:378: :hackney.send_request/2
    (httpoison 1.8.0) lib/httpoison/base.ex:846: HTTPoison.Base.request/6

The issue seems to be from hackney not supporting status codes as floats and is actively trying to convert 401.1 into an integer. Does anyone have a workaround for this?

looks like there are some float status codes here.

2

The HTTP/1.1 spec, RFC7230 specifically defines the status code as 3 characters from 0 to 9.

The server in your post returns the following response (curl session):

*   Trying 103.211.70.61...
* TCP_NODELAY set
* Connected to www.lowiqcanadian.com (103.211.70.61) port 80 (#0)
> GET /id2.html HTTP/1.1
> Host: www.lowiqcanadian.com
> User-Agent: curl/7.64.1
> Accept: */*
> 
< HTTP/1.1 401.1 Unauthorized
< Server: nginx
< Date: Tue, 09 Nov 2021 00:19:45 GMT
< Content-Type: text/html; charset=utf-8
< Transfer-Encoding: chunked
< Connection: keep-alive
< 
<html lang="en"><head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
...Chinese linkfarm content omitted...
3

That’s a helpful reference, thanks! Do you think an HTTP client should only tolerate responses that follow the spec? Or should they try to accommodate the fact that some servers will bend/break rules? I.e. do you think there is any point in submitting a PR for the related issue Status codes as floats · Issue #697 · benoitc/hackney · GitHub ? Or is that just opening a can of worms?

4

I asked this question on stackoverflow to see what comes back. Seems very strange that iIS would return non-compliant status codes and I can’t find any update to RFC7231 that would suggest the standard has been updated.

1 Like
6

gotcha, will monitor the responses on there also.

7

The response in stack overflow indicated that IIS only logs the extended status code but that the HTTP header is in fact an integer.

The example above is from an nginx server but I can’t find any reference to float status codes there. Some now I’m even more curious.

8

And searching the nginx repo mirror doesn’t seem to find the string 401.1 either.

9

I think the site was built using Tengine (based off nginx), but I don’t see any references to fractional status codes there either.

10

My interpretation of those response headers is that there’s an nginx-based load-balancer in front of an IIS server.

Alternatively, it could be an IIS server deliberately obfuscating its identity to confound IP-space scanners.

11

It is a badly configured junk site for sure, but still Hackney should not crash. I tried a few http clients in other languages and most of them give some sort of degraded response instead of crashing.

1 Like