Hello Andy,

Thank you for your review.

We've adopted several of your suggestions in:
https://github.com/mxsasha/nrtmv4/commit/5d4918454b3fcc3902ccb797f11d33bb58e82c5c
More responses inline:

On 10 Apr 2026, at 01:02, Andy Newton via Datatracker <[email protected]> wrote
> 315        server publishing NRTMv4 for RIPE and RIPE-NONAUTH, will generate 
> two
> 
> Would "RIPE and RIPE-NONAUTH (two separate IRR databases)" make it clearer
> to the reader that RIPE and RIPE-NOAUTH refer to separate databases?

Yes, that would be clearer. Adopted.

> 416        ...  It is RECOMMENDED to
> 417        modify objects in such a way that this change is evident to humans
> 418        reading the object text, for example, by adding remark lines or
> 419        comments.
> 
> Without a specific action to take, is this RECOMMENDED normative? Can the
> advice to add remarks or comments be made THE action to take?

This RECOMMENDED is meant to refer to the action of marking a server-modified 
object as such. We changed the language to make this clearer.

> 934        It is RECOMMENDED for mirror clients to be flexible where possible 
> [..]
> 
> This is good advice, but it is unclear what action is to be taken to be
> flexible. I think this should be a non-normative "recommended".

I agree, this is now non-normative.

> 982        It is RECOMMENDED that IRR Database operators rotate the signing 
> key
> 983        on their mirror server about once per year.  ....
> 
> The "about" leaves a lot of room for interpretation. Can a range be specified,
> such as between 8 months and 16 months?

We can, but I do not feel that is a real improvement. The range is 
intentionally vague, the intended boundaries are: not so often that it becomes 
annoying; not so rarely that it becomes an unfamiliar procedure.

> Has the working group considered using media types to describe the type of
> content in the files, instead of relying on file name suffixes? That might
> offer some flexibility for switching to different compression types, etc..., 
> in
> the future.

This has not come up before. It's a valid point, but we feel this change would 
be too impactful at this time.

> I know this is a nit, but I think the objects are "respresented using RPSL",
> where storage could be done by any means... such as a denormalized SQL db.

Entirely agreed. We changed it to represented.

Thanks,
Sasha
_______________________________________________
GROW mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to