[
https://issues.apache.org/jira/browse/QPID-640?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12534844
]
Rafael H. Schloming commented on QPID-640:
------------------------------------------
A good example to think of here is the content-length field. A content-length
of zero is not the same thing as no content-length at all.
> The C++ struct encode/decode always includes nested structs even when they're
> empty, and (I believe) transforms empty strings to null
> -------------------------------------------------------------------------------------------------------------------------------------
>
> Key: QPID-640
> URL: https://issues.apache.org/jira/browse/QPID-640
> Project: Qpid
> Issue Type: Bug
> Components: C++ Broker
> Reporter: Rafael H. Schloming
>
> The code that detectes whether or not a struct field is present does so by
> checking the size or value of the field. This results in some odd behavior.
> - nested structs will always be included on the wire even when empty because
> the size is computed to always include the packing flags
> - empy strings can never be encoded onto the wire
> - zero values will be omitted
> As the specification doesn't state that empty values are equivalent to absent
> ones, this behavior is probably unsafe since to clients that treat null
> separately from empty ("" or 0), the broker would appear to be transforming
> the specified field value.
> As a general rule I'd suggest that when the client presents us fields encoded
> in a certain way we should probably leave the encoding of those fields alone
> unless we're modifying the semantic value somehow.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
