Re: [quicwg/base-drafts] encoding of CONNECTION_CLOSE reason phrases (#1990)

[Kazuho Oku <[e-mail]>]

@mikkelfj 
> If it's UTF-8 it is much simpler to dump in error logs, even if you don't bother validate it. Some logs tend to use JSON which is practically always UTF-8.

Don't do that. Because this is how you create a DoS attack vector on the server side.

If the reason phase contained an invalid UTF-8 sequence, creating a JSON that contains that reason phase makes the entire JSON invalid. Because many JSON decoders refuse to return anything when the input contains an invalid UTF-8 sequence, you essentially destroy all the information that were stored in JSON.

For example, consider you have an automated system that collects the reason phases for all the connections sent by the peer. Having just one invalid reason phase stops the entire system.

To summarize, calling it "MUST UTF-8" does not make things simpler.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/1990#issuecomment-437856770