|
»
« Return to Thread: Erratum on RDDP Architecture RFC
Re: Erratum on RDDP Architecture RFC
by Talpey, Thomas
::
Rate this Message:
At 04:49 PM 12/1/2008, Black_David@... wrote:
>A pair of sharp eyes caught a couple of related issues
>in the RDDP Architecture RFC:
>
>http://www.rfc-editor.org/errata_search.php?eid=136
>
>In summary, the first prototype for rdma_read is:
>
> rdma_read(socket_t s, ddp_addr_t s, ddp_addr_t d);
>
>and the second prototype doesn't match:
>
> rdma_read(socket_t s, ddp_addr_t s, length_t l, ddp_addr_t d)
>
>plus the second prototype has two arguments that are both
>named "s". Oops ...
>
>I believe that the correct fix is that the length argument
>from the second prototype needs to be in both prototypes,
>and the DDP buffer argument "s" (source) should be renamed
>to "rs" (remote source) in both prototypes in order to
>avoid a name conflict with the socket argument.
>
>If anyone disagrees, please respond.
Yes, the authors agreed on the first matter when it was pointed out
to us back in 2006.
The first prototype is indeed missing a "length". The second prototype is
correct, and the first needs to be changed, per the existing rfc-editor note.
The second change, however, is unnecessary. As the authors pointed out in
our original response:
"As for the duplicated argument name, this is a minor issue. As the
document states in section 1 (Introduction):
"This document uses C language notation as a shorthand to
describe the architectural elements of DDP and RDMA protocols. The
choice of C notation is not intended to describe concrete protocols
or programming interfaces."
Because there are no further references to these arguments by
name, this apparent inconsistency is merely a detail. The text
would have the same meaning without the "s" parameter names
altogether."
Our position that this second issue is not an erratum, has not changed.
The first should in any case be executed. Has the RFC editor been waiting
for some confirmation from the Working Group or authors on #136? We would
be happy to give the latter, if so.
Tom.
>
>Assuming this approach is ok, I propose to carry it out
>as follows:
>- I will file a new erratum with the above two changes
>- I will then ask the RFC Editor to reject the original
> erratum, as it contains a correction for only one
> of the two problems that it identifies.
>
>Thanks,
>--David
_______________________________________________
rddp mailing list
rddp@...
https://www.ietf.org/mailman/listinfo/rddp
>A pair of sharp eyes caught a couple of related issues
>in the RDDP Architecture RFC:
>
>http://www.rfc-editor.org/errata_search.php?eid=136
>
>In summary, the first prototype for rdma_read is:
>
> rdma_read(socket_t s, ddp_addr_t s, ddp_addr_t d);
>
>and the second prototype doesn't match:
>
> rdma_read(socket_t s, ddp_addr_t s, length_t l, ddp_addr_t d)
>
>plus the second prototype has two arguments that are both
>named "s". Oops ...
>
>I believe that the correct fix is that the length argument
>from the second prototype needs to be in both prototypes,
>and the DDP buffer argument "s" (source) should be renamed
>to "rs" (remote source) in both prototypes in order to
>avoid a name conflict with the socket argument.
>
>If anyone disagrees, please respond.
Yes, the authors agreed on the first matter when it was pointed out
to us back in 2006.
The first prototype is indeed missing a "length". The second prototype is
correct, and the first needs to be changed, per the existing rfc-editor note.
The second change, however, is unnecessary. As the authors pointed out in
our original response:
"As for the duplicated argument name, this is a minor issue. As the
document states in section 1 (Introduction):
"This document uses C language notation as a shorthand to
describe the architectural elements of DDP and RDMA protocols. The
choice of C notation is not intended to describe concrete protocols
or programming interfaces."
Because there are no further references to these arguments by
name, this apparent inconsistency is merely a detail. The text
would have the same meaning without the "s" parameter names
altogether."
Our position that this second issue is not an erratum, has not changed.
The first should in any case be executed. Has the RFC editor been waiting
for some confirmation from the Working Group or authors on #136? We would
be happy to give the latter, if so.
Tom.
>
>Assuming this approach is ok, I propose to carry it out
>as follows:
>- I will file a new erratum with the above two changes
>- I will then ask the RFC Editor to reject the original
> erratum, as it contains a correction for only one
> of the two problems that it identifies.
>
>Thanks,
>--David
_______________________________________________
rddp mailing list
rddp@...
https://www.ietf.org/mailman/listinfo/rddp
« Return to Thread: Erratum on RDDP Architecture RFC
| Free embeddable forum powered by Nabble | Forum Help |
