Re: [PHP-DEV] Re: Literal / Taint checking

This is only part of a thread. view whole thread
March 11, 2020 13:09 (Craig Francis)
On Mon, 9 Mar 2020 at 16:54, Rowan Tommins> wrote:

> [...] the way to get a more in-depth discussion going is to draft an RFC
Thanks Rowan, I've created a Wiki account (craigfrancis), and I believe the next step is to ask for RFC karma? And is there is anyone who can help with the technical details? I'd really appreciate it. If it helps, I've got a development budget of £1,000 I could put towards this (I will need an invoice, a simple PDF would do). As to the discussion around the older RFC[1]... The last message I can find was on the 1st September 2015[2]. The RFC focused on SQL injection, where it was noted that "unfiltered input can affect way more than only SQL"[3], and it isn't ideal for "just for one use case"[4] - my proposed `is_literal()` can be used for other issues, such as Cross-Site Scripting[5], Command Line Injection, etc. There was a belief that education was the answer[6] - but having this check would allow developers to identify (and block) mistakes at runtime. Xinchen mentioned how it was complex in PHP5 to implement the Taint extension - but "with PHP7's new zend_string, and string flags, the implementation will become easier"[7]. And while the Taint checking is useful, it does not address the mistakes that can happen with escaping. As to why I'm deviating away from the original RFC... By providing a `is_literal()` function, it allows the developer to determine how they want to use it - where they can skip it for certain tasks[8], and database drivers (or other extensions) could use it in the future to raise a notice/warning/error[9]. It gives a mechanism to ensure inputs are split between the command (a literal), and user supplied values - which is what Yasuo was asking for[10].. Also, by focusing on just literals (as in, only values defined within PHP scripts), we avoid any concerns about escaping (which can go wrong), and we won't need to identify which sources are trusted[11]. For the last 5 years I've been writing my SQL with literals only, and it's worked very well... with just one oddity (which I still consider a literal): [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] On Mon, 9 Mar 2020 at 16:54, Rowan Tommins> wrote:
> On Mon, 9 Mar 2020 at 13:47, Craig Francis <> > wrote: > > > Hi, > > > > As I'm not sure how to make any more process on this, I've added added a > > Feature Request: > > > > > > > > It shows how this change in PHP could stop SQL injection, and proposes a > > way it could be used against HTML injection as well. > > > > > Hi Craig, > > In my experience, the bug tracker is likely to get you less attention than > this list, rather than more. For this kind of significant change, the way > to get a more in-depth discussion going is to draft an RFC; there are some > instructions and tips on how to go about that at > and > > > > The idea of an RFC is to sit down and design exactly how the proposed > feature would work; that helps move the discussion forward, because people > can see exactly how it might look, and means you're offering something to > the community rather than asking it of them. The RFC doesn't have to > include a full implementation, but if you don't know much about the > technical details, you might need help from someone who does to make sure > the proposal is realistic. > > I see you've linked an older RFC in the feature request; it would be worth > digging out the archived discussion from when that was proposed, to see why > it stalled. It may just be that people were distracted by other things, or > there may be issues raised which you can consider in a new proposal. If you > haven't already, you could try contacting the author as well. > > In general, I think it's an interesting idea, but as the saying goes "the > devil is in the detail", so I don't have much to say without a concrete > proposal for what it would look like. > > Regards, > -- > Rowan Tommins > [IMSoP] >