[PATCH 0 of 1] Bad input handling: discussion

[Mads Kiilerich]

On 02/26/2015 09:40 PM, Andrew Shadura wrote:
> Hello,
>
> On Thu, 26 Feb 2015 21:06:36 +0100
> Thomas De Schampheleire
> <patrickdepinguin at gmail.com> wrote:
>
>> I don't think Kallithea should crash or present 500 in cases where an
>> input is not what we expect.
>> In any case, '400 bad request' is better than '500 server error', as
>> also said by Mads in [1].
>>
>> Whether or not we should ignore invalid input: my initial thought was
>> that it is good idea. However, from the link Mads provided in [1], it
>> seems there can be security issues with such behavior, in general. So
>> I'm not sure anymore what to do here, I'm not very familiar with this
>> area.
>>
>> What could be the reason for such invalid input, other than malicious
>> attempts?
> In this situation (I found "WS%3" in the real logs) this might be
> misinterpretation of some links by search engine bots or something like
> that. I'm not sure how failing on such input is better or worse than
> ignoring it.

Yes, that is where I prefer "garbage in - garbage out". We could try to 
behave "nicely" when we get "garbage in", but there is really no correct 
response to "garbage in". I prefer to make it clear that we can't 
respond correctly instead of "cheating" and trying to guess what the 
intention was ... and thus making the user believe the output is 
reliable when it isn't.

/Mads