Skip to contentSkip to conversation

Follow up on "call helper function by address" terminology

DISCUSSION [email protected] 4 participants 6 messages Apr 5, 2024 – Apr 18, 2024 BPF

lore
Share & tools
lore
#0 lore raw

At IETF 119, we agreed that "by address" should be changed to something else in the ISA.
The term "legacy ID" was used during the discussion but Christoph (if I remember right)
pointed out that such IDs are not deprecated per se. Hence "legacy" may not be the
right word since we use that word with legacy packet access instructions that are
deprecated. We decided to take further discussion to the list, hence this email.

We need some term to distinguish them from BTF IDs, so another alternative might be "non-BTF ID".

Thoughts, or other alternatives?

Dave

#1 lore raw
Quoted 10 lines · On Fri, Apr 05, 2024 at 01:10:38PM -0700…

At IETF 119, we agreed that "by address" should be changed to something else in the ISA. The term "legacy ID" was used during the discussion but Christoph (if I remember right) pointed out that such IDs are not deprecated per se. Hence "legacy" may not be the right word since we use that word with legacy packet access instructions that are deprecated. We decided to take further discussion to the list, hence this email.

We need some term to distinguish them from BTF IDs, so another alternative might be "non-BTF ID".

Non-BTF ID is fine with me. Any objections?

Thanks,
David

#2 lore raw
Quoted 10 lines · David Vernet
Quoted 8 lines · On Fri, Apr 05, 2024 at 01:10:38PM -0700…

At IETF 119, we agreed that "by address" should be changed to something else in the ISA. The term "legacy ID" was used during the discussion but Christoph (if I remember right) pointed out that such IDs are not deprecated per se. Hence "legacy" may not be the right word since we use that word with legacy packet access instructions that are deprecated. We decided to take further discussion to the list, hence this email.

Quoted 4 lines
Quoted 2 lines

We need some term to distinguish them from BTF IDs, so another alternative might be "non-BTF ID".

Non-BTF ID is fine with me. Any objections?

If something later comes along supplanting BTF it will be the not-BTF not-non-BTF thing. This is bad. How about untyped identifiers?

Quoted 7 lines

Thanks,
David
--
Bpf mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/bpf

Signature · 1 line
Astra mortemque praestare gradatim
#3 lore raw
Quoted 11 lines · Watson Ladd
Quoted 1 line · David Vernet

On Fri, Apr 05, 2024 at 01:10:38PM -0700,

Quoted 8 lines · [email protected] wrote:
Quoted 8 lines

At IETF 119, we agreed that "by address" should be changed to something else in the ISA. The term "legacy ID" was used during the discussion but Christoph (if I remember right) pointed out that such IDs are not deprecated per se. Hence "legacy" may not be the right word since we use that word with legacy packet access instructions that are deprecated. We decided to take further discussion to the list, hence this email.

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

We need some term to distinguish them from BTF IDs, so another alternative might be "non-BTF ID".

Non-BTF ID is fine with me. Any objections?

If something later comes along supplanting BTF it will be the not-BTF not-non- BTF thing. This is bad. How about untyped identifiers?

For runtimes that have a way to look up type info from a non-BTF ID, the ID is not "untyped" per se.

Other possibilities:
* Classic ID, but "classic" would imply classic BPF
* Index, but that would imply the runtime actually has to implement it as an index

As such, I think "non-BTF ID" is better than the other possibilities above, and a future ISA version could always rename it if other things come up in the future that necessitate a terminology change.

Dave

#5 lore raw
Quoted 1 line · Christoph Hellwig

Maybe "static ID", or "pre-assigned" id?

I'd be ok with either of those. If I don't hear
otherwise, I'll create a patch using "static ID".

Dave