Re: pkgs contain non URL safe characters

Just to check this behaviour, I used tcpdump to see what the request 
looked like from pkg-fetch.


     123.ish.com.au.15580 > pkg0.twn.freebsd.org.http: Flags [P.], cksum 
0x80e0 (incorrect -> 0xfc82), seq 1:184, ack 1, win 1027, options 
[nop,nop,TS val 975600196 ecr 3136747760], length 183: HTTP, length: 183
     GET /FreeBSD:13:amd64/quarterly/All/openjdk11-11.0.13+8.1.pkg HTTP/1.1
     Host: pkgmir.geo.freebsd.org
     Accept: */*
     User-Agent: pkg/1.17.5
     Range: bytes=6733824-
     Connection: close


You can see in there that the + is not URL encoded. Is it expected that 
pkg uses URL standards for its repository? If not, any advice on how to 
host a repository on a commercial service like AWS cloudfront?

Should we rewrite all our files with + symbols to spaces? Should pkg 
names only contain URL safe characters? Or should pkg-fetch be fixed to 
encode URLs?


I took a quick look at the source for pkg.c and where it calls fetchXGet 
but I can't understand where any URL encoding might happen.


Ari


On 14/2/2022 11:18am, Aristedes Maniatis wrote:
> Some packages contain "+" symbol which is a way of encoding spaces in 
> a URL. This means that I'm having trouble hosting our pkg repository 
> behind cloudfront/S3.
>
> I wasn't sure where to post this issue, so I put more details here: 
> https://github.com/freebsd/poudriere/issues/976
>
>
> Is there a workaround for this issue? Could pkg-fetch escape such 
> characters when interacting with a http repository?
>
>
> Cheers
>
> Ari
>
>