Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

Loading…

A.1 should include HoTT features #74

Closed
cangiuli opened this Issue · 20 comments

7 participants

Carlo Angiuli Mike Shulman Daniel R. Grayson Steve Awodey Andrej Bauer petera02 coquand
Carlo Angiuli
Collaborator

The presentation of type theory in A.1 (of Peter and Thierry) does not mention univalence, definitional eta, or higher inductive types.

petera02 petera02 was assigned
coquand coquand was assigned
Mike Shulman
Owner

A.2.15 mentions funext and univalence, A.2.12 does definitional eta for functions, and A.2.14 does a HIT, so I think this is done.

Daniel R. Grayson
Collaborator

Except Carlo opened it before the intro lost its number, so now the issue should be considered as referring to A.1.

Daniel R. Grayson DanGrayson reopened this
Daniel R. Grayson
Collaborator

Peter Aczel wants to deal with the issue by adding a statement that A.1 doesn't include any HoTT features.

Steve Awodey
Owner
Andrej Bauer
Owner

But there is no point in repeating these twice. Let us just promote A.2.15 to A.3, i.e., make it its own section. Say in the first sentence of the section that these are given in the style of the second presentation. Actually, I am just going to do this.

Andrej Bauer
Owner

Closed by b48deb7
This looks better. The table of contents now has a line saying "The axioms for HoTT" so people can actually like them.

Carlo Angiuli
Collaborator

I don't quite like this solution because the presentation in "Axioms for HoTT" is dissimilar from the type theory presented in "The first presentation."

More importantly, "The first presentation" doesn't mention HITs. Since it isn't obvious to me the best way to integrate HITs into the first presentation (since elimination is done via pattern matching) I think Thierry or Peter need to make some remark about this.

Carlo Angiuli cangiuli reopened this
Steve Awodey
Owner
Steve Awodey
Owner

there can be a remark in the "intro" part that says the two sections of the appendix treat the "basic" theory in two different styles, and then the HoTT axioms are given in the third section, in the style of section 2. No reason to complicated things any further.

Carlo Angiuli
Collaborator

@awodey: Sure, I'll do that. But I still think HITs should be discussed separately in each section?

Steve Awodey
Owner

why?
It's not necessary to do it twice (or three times), and Peter Aczel and Thierry have already said they prefer not to do it in theirs. So why not just do it once, in the third section, and not in the other two?

Carlo Angiuli
Collaborator

See 911d119

I have added a sentence to clarify that the UA/funext presentation of A.3 works equally well for both A.1 and A.2.

I am satisfied now, but I'd just like to remark that the suitability of pattern-matching for HITs is not so clear, even to experts. If we don't want to broach the issue in A.1, that's fine, but it would be unfair to say (and the appendix currently does not say) that HITs also work equally well in both sections. Therefore, please do not move the circle to A.3.

Carlo Angiuli cangiuli closed this
Mike Shulman
Owner
Carlo Angiuli
Collaborator

Yes, that was the kind of thing that I was hoping would be added to A.1, except that it would invalidate the claims of A.1.9, and so requires more work than just pasting in that text. But I get the impression that Peter and Thierry don't want the HoTT features in their section because it screws up the syntactic properties. (I disagree with this reasoning but that is why the issue is assigned to Thierry.)

Mike Shulman
Owner

Well, if that's the point, then we should move A.2.14 into A.3: if we're claiming that the two presentations are equally good, then all the special HoTT features should treat them equally, either being discussed separately in both A.1 and A.2, or being discussed after both of them in A.3.

Carlo Angiuli
Collaborator

But (1) A.2 doesn't have any operational semantics, unlike A.1, and (2) it's a bit more of a stretch to say that A.2.14 can be dropped into A.1 without further discussion.

I guess my point is that A.2+A.3 forms a coherent presentation of the HoTT we use in the book, while A.1+A.2.14+A.3 is not so internally consistent.

Mike Shulman
Owner
Carlo Angiuli
Collaborator

The core feature of A.1 is that their notion of definitional equivalence is inherited from untyped operational semantics. Without that, A.1 is basically a reprisal of Chapter 1 with slightly more formal wording.

Maybe it would be easier to discuss this offline tomorrow. In principle I don't mind entirely factoring out the HoTT features as you suggest.

Andrej Bauer
Owner

Let's not get too emotionally attached to who wrote what (i.e., delete one of the section if it doesn't fit). But I think it's good to have the two presentations.

We should place all the HoTT features into a separate section, as far as I am concerned. This makes sense from the organizational point of view. For example, Martin Escardo asked that we do such a thing, so he can have a look at what is different in the book from the standard setup. He's not going to care whether it is done twice, or in this or that style, he just wants to have one place where he can look at the special HoTT stuff. We can always explain (and should) in the notes how the special HoTT stuff relates to A.1 and A.2.

Carlo Angiuli
Collaborator

I opened #128; I'll make some changes and then we can continue the discussion there.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.