Re: http-client port encoding failed

[tomas]

On Fri, Sep 04, 2020 at 02:35:45PM -0500, mbcladwell@stihie.net wrote:
> 
> Hi,
> 
> I am using http-client to submit requests that contain extended
> Latin characters e.g. "Marjanović+Ž".  I believe I have locales set
> up properly as I don't receive the "can't set locales" error.  Also
> I can regex with the extended Latin characters.
> 
> scheme@(guile-user)> (locale-encoding)
> $11 = "UTF-8"
> 
>  When I submit a url containing the above Serbian name, I get the error:
> 
> Throw to key `encoding-error' with args `("put-char" "conversion to
> port encoding failed" 84 #<input-output: string 7febcf44c2a0> #\ć)'.

I'm not sure and can't double-check at the moment, but: the HTTP header
part (and the URL is part of it) /must/ be 7 bit ASCII. That's why there
is that URL encoding [1] -- basically you take the UTF-8 encoded string
and represent every byte (octet) beyond 0x7f (and some "hot" bytes
beneath that) by %<xx>, where <xx> is the hexadecimal representation
of that byte.

Perhaps you have to URL-encode your URL before "submitting" it.

> If I paste the URL into my browser I receive the correct response.

Your browser does all of the above for you. It just doesn't tell you :)

> Prior to the request I use (set-port-encoding! (current-output-port)
> "UTF-8") without effect.

Hm. I don't know what the "right" encoding for a HTTP connection
is. But the HTTP header part is 7 biz ASCII (which is a subset
of UTF-8, but just that :)

The quintessence is: try url-encoding your URL.

Cheers
 - t

signature.asc
Description: Digital signature