Removal of smallvec from dependancies.

[darnuria]

Hello, taking interrest with issue: #365

Disclamer: the approach is naive I know it involve a most to malloc and leaved the code untouched even allocating in a loop, mostly to began exploring the subject! :)

In crate/resolver I chosed to use a plain vec since the type: SmallVec::<[Nameservers<P>; 2] reached 544 bytes not sure it's small anymore. (did a little std::mem::size_of::<SmallVec<[NameServer<P>;2]>>() to get the size in bytes)

Why I didn't used TinyVec directly? Because it would have needed a Default implementation on each types of Nameservers.

If it's not the best course of action I am ok to get some mentoring! :)

Some idea I have in mind:

• If it's not the right course of action: either close Evaluate all Vec usage and replace with SmallVec where apropriate #365, or reformulate it and document here why smallVec is mandatory here.
• If it's OK but need tinyVec, will need some work since Default is mandator due to TinyVec Array trait.
• If the plain Vec is good enough then just have something less naive ! ;)

[bluejekyll]

Sorry for the late response here. This looks really good to me. I like simplifying the dependency graph, and SmallVec is probably not buying much here. For reference, the idea here was that since the vast majority of use cases would probably not override this setting (2 NameServers for lookup) then it would in theory be more efficient to store a fixed size for that use case. This might not be well-founded, and as you already identified, may not actually save us much in terms of allocation.

[bluejekyll]

Do you mind if I update this branch for you? or would you like to rebase?

[bluejekyll]

@darnuria, I just noticed this was a draft, feel free to turn this into a PR ready for review, I'll approve and merge.

[darnuria]

Will work on it today or tomorrow ! No problem it's floss/libre rules no pressure it's OK to have some work stalling didn't have a contract with you! :D So best effort tomorrow for me to rebase/update! Feel free to review, I viewed that tiny-vec tries to move away from the Array trait maybe it change the game a little bit for us. (will add the info I saw tomorrow)

…

On 7/27/23 18:51, Benjamin Fry wrote: Do you mind if I update this branch for you? or would you like to rebase?