1. This should be incorporated into com_joomlaupdate as well so core can use it.

2. We need to make a decision on what hashes to support. In 2017, md5 and sha1 aren't great options.

We need to make a decision on what hashes to support. In 2017, md5 and sha1 aren't great options.

sha256 ?

I can't say I know much about sha256's security off hand.

This should be incorporated into com_joomlaupdate as well so core can use it.

sure, can we do in a separate pr ?

We need to make a decision on what hashes to support. In 2017, md5 and sha1 aren't great options.

sha256 should be OK nowdays ...or not 😕

I already know what he's going to nudge toward but I'm going to ask an opinion anyway.

@paragonie-scott are there any major issues to be aware of with using sha256 or sha512 with hash_file() as a means for validating a checksum of an extension (or our core) package as a short-to-mid term solution? (Clearly we could/should continue to work for better options, so right now this is basically a starting point; the one thing this wouldn't address is a compromise of anyone's distribution platform)

in the gsoc 17 EEM meeting yesterday, i've discovered that there are some 3dp extensions that use
md5/sha1 like ARS, so apart the discussion about weakness, should we consider to support those weak algos?
one step forward (sha512) and one step back (md5,sha1)

I'm well aware that ARS generates them. That still doesn't necessarily make md5/sha1 good to introduce into new APIs in the core libraries in 2017 as a means for validating package integrity.

are there any major issues to be aware of with using sha256 or sha512 with hash_file() as a means for validating a checksum of an extension (or our core) package as a short-to-mid term solution?

There's a length extension attack, but that doesn't break preimage security, and since it's an unkeyed hash, it's not giving the attacker a capability they don't already have.

However, if you want to be conservative here, follow Composer's example and go with SHA384 specifically.

so we can go with: sha256,sha384,sha512 ?

That seems fair to me.

now hash supported are:

• sha256
• sha348
• sha512

I have tested this item ✅ successfully on 2f3425f

Tested OK, as part of #17632

For more info about the test and comment: https://issues.joomla.org/tracker/joomla-cms/17632#event-3

I have tested this item ✅ successfully on 2f3425f

RTC after two successful tests.

@wilsonge Can we get this and #17632 merged into the 3.9 branch? IMO now would be a good time for the code to hit core and start documenting the hell out of the new functionality via the docs wiki and a targeted blog post on the dev network.

LGTM :)

Well bugger - didn't see this was against staging. Reverted with #17619 and going to manually re-patch against 3.9 now

Merged second time lucky with f18c20a

@wilsonge thanks
can you consider to evaluate/merge to 3.9 #17632 too after conflict has been resolved ?