My thoughts:

1. We should avoid the use of innerHTML in framework code as much as possible.
2. We should discourage the use of it in application code, although I think 
it’s not going to be very heavily used in general in Royale no-matter.
3. We are using innerHTML in parsing MXML and innerHTML is one of the default 
MXML properties. This seems to be safe as it's compile-time.

Harbs

> On Mar 16, 2020, at 7:17 AM, Greg Dove <[email protected]> wrote:
> 
> Before I forget, I believe we need to have a discussion about the use of
> innerHTML inside the framework before we get to 1.0 release. A few of us
> discussed this briefly in Slack a couple of months back, and we decided it
> needed shared input/scrutiny on list. I'm only raising it here to achieve
> that.
> 
> Replying to this
> I'm suggesting this needs attention, but I don't have any simple answer.
> Please share your thoughts/ideas.
> 
> Why does it need scrutiny?
> While I can't see a direct security risk (it seems the main risk is in
> combination with other things), it is considered at least a weak point, and
> it sounds like it would probably cause a fail in a security audit of an
> application. Aside from that, if there is real risk then it could affect
> Royale's reputation, or that of Apache.
> I think we are using innertHTML at various depths in a number of different
> places from a quick 'search in files'. If values for setting these could
> come from externally created strings that are not filtered for security
> risks (e.g. <script> tags), then I think those would be the areas of
> concern in a security audit.
> 
> Others
> I already know what React does for this, but I did not look elsewhere, so I
> don't know what other frameworks do. With React they make it possible to
> use innerHTML, but doing so in itself is a very clear reminder that it is
> not a great thing to do.
> See here :
> https://reactjs.org/docs/dom-elements.html#dangerouslysetinnerhtml

Reply via email to