[RFC] deprecate md5_file and sha1_file

I disagree. While MD5 and SHA1 might not be suitable for modern cryptographic operations, these functions might be needed for legacy situations -- e.g., munging through old data. On Mon, Feb 10, 2020 at 1:50 PM Tom Van Looy via internals < internals@lists.php.net> wrote:

On Mon, 10 Feb 2020 at 22:50, Tom Van Looy via internals < internals@lists.php.net> wrote:

I feel that if we deprecate the file versions of these algorithms we probably should also deprecate the ones which work with plain old strings, namely md5() and sha1(). [1] [2] It should be noted that these hash functions would still be available through the Hash extension which is always available as of PHP 7.4 as it became a core extension. Depending on how controversial these deprecations are they could be bundled with the big deprecation RFC for PHP 8.0, which is currently in draft. [3] George P. Banyard [1] https://www.php.net/manual/en/function.md5.php [2] https://www.php.net/manual/en/function.sha1.php [3] https://wiki.php.net/rfc/deprecations_php_8_0

On 10/02/2020 21:49, Tom Van Looy via internals wrote:

It won't. At best it will make people switch to the hash function. At worst people will not upgrade. If people are using the existing md5 / sha1 algorithms, chances are it's because they're actually wanting to get a hash to compare to something that has already been stored. There's not much point in deprecating the algorithm if we don't eventually plan to remove it, and there is an exactly zero percent chance of it being removed at any point in the next 50 years. Mark Randall

Hi!

Not sure what's the point of it. SHA1 and MD5 didn't change. The recommendations for their usage has changed, but we generally don't use deprecation warnings to improve users' code and recommend them how to structure their code better. Deprecation is usually for functions that are unsupported (and going to go away) or not having any non-broken uses anymore. We have crc32() function which is much less secure than md5 for check-summing obviously, and we don't deprecate it because of that. I don't think we should use deprecations for that purpose.

There's zero effort expended on "carrying around" these functions, and I don't see why it's "too much" for "modern PHP"? What changed in "modern PHP" that suddenly brought forward the need to minimize the number of existing functions? I don't see any reason for that. On the contrary, deprecating it causes work for people that legitimately use these functions, and makes them either turn off deprecation warnings (unsafe and may lead to real deprecations pass unnoticed) or work around this. I don't think it is beneficial.

As others have mentioned, this will do nothing but make people annoyed and switch to the hash_file() version of exactly the same thing or put up another hurdle to updgrading PHP. The password hashing API now provides an obvious go-to for password hashing. For other hashing usages there are, I think, basically two scenarios developers find themselves in: 1) I'm using an API or some other external service and that requires the use of md5 / sha1 - I don't have a choice 2) What do I use instead? Internet searches return "sha1 / md5 is fine for this purpose" or recommend algorithms that aren't natively supported in PHP. The (hash library) documentation does nothing towards helping developers decide what algorithms (or even which hash library functions) they should use for what purposes (and there are a lot of acronyms that many developers are likely to have never encountered that are never explained - HMAC, PBKDF2, HKDF). Yes, of course developers can use third party sources to supplement the information in the manual, but who has time to go seartching for that (esp. when most of the first page on Google probably tells you md5/sha1 is fine anyway)? If you want to change the way developers think about hashing when writing PHP, I would start with the documentation rather than deprecating functions which are essentially aliases and are highly likely used all over the place in cases where they do exactly what people want. AllenJB On 10/02/2020 21:49, Tom Van Looy via internals wrote: