I think he's referring to this discussion: https://mail.openjdk.org/pipermail/amber-spec-experts/2024-A...

Thanks! We've changed the top link to that from https://bugs.openjdk.org/browse/JDK-8329949, which is mostly just a pointer.

It's not just you. My reading is that the flaw was toxic Java users made the developers feel so bad about the feature that they canceled it out of spite. I'm almost certain that's not what it was but that's the only information in there.

Yeah lots of bikeshedding over \ instead of $ was just really unnecessary.

This might be a better post. Brian's original email is here.

https://mail.openjdk.org/pipermail/amber-spec-experts/2024-A...

> In the course of using this feature in the `jextract` project, we did learn quite a few things we didn’t already know, and this was conclusive enough that it has motivated us to adjust our approach in this feature. Specifically, the role of processors is “outsized” to the value they offer, and, after further exploration, we now believe it is possible to achieve the goals of the feature without an explicit “processor” abstraction at all! This is a very positive development.

My Summary:

- nothing to do with the choice of backslash vs dollar. Doubles down on dollar being bad (due to backward and cross compat with other languages)

- original choice to use "Template Processors" misguided, reverses course to StringTemplate Literals

- this requires a new paradigm where APIs now need to explicitly support StringTemplates, instead of relying on the Processor to convert them into Strings as necessary (big leap in my logic here)

- StringTemplates are explicitly still not Strings. This is in line with their "Security Focus First"

- sensitive APIs that depend on StringTemplates should explicitly treat their interpolated results carefully through overloads in a sensitive context.

Obviously I don't know much about the `jextract` project but I can guess 1 possible way this breaks the security concept is if you use the wrong Processor to pass into a Db.query method, for example.

    db.query(STR."select * from users where id = \{id};");

This way, you can still do it, but its just much harder to do. Or as an API developer you could deprecate your String overload and insist only on StringTemplate usage from now on.

    db.query("select * from users where id = \{id};"); // this is compiled as a StringTemplate

    db.query("select * from users where id = \"" + id "\""); // this is still a String, and it's still bad, but can be mitigated by deprecating Stringm method

Disclaimer: i haven't done java in nearly 10 years... i just love theorycrafting. So take my contribution with a pinch of salt please :)

That's actually a really clever mechanism. Although it'll make the common case (basic string interpolation, no need for security concerns) more complicated unless a lot of APIs support a StringTemplate overload.

This could be supported with String constructor with StringTemplate overload.

    String str = new String("Hello \{name}");

OR just add a .interpolate() function

    String str = "Hello \{name}".interpolate(); // verbose
    String str = "Hello \{name}".join(); // less verbose, but less representative

OR probably at the bottom of the list of recommendations

    String str = "Hello \{name}".toString() // matches StringBuilder behaviour

Knowing Java devs, also a non-zero chance it becomes

    String str = "Hello \{name}".toUnsafeString()

Quoting Brian:

> shed to be painted (but please, not right now.).