BIND 10 #3263: RRset::addRdata(ConstRdataPtr&) does not validate RDATA against RRset's class (was: RRset::addRRset(ConstRdataPtr&) does not validate RDATA against RRset's class)

[BIND 10 Development]

#3263: RRset::addRdata(ConstRdataPtr&) does not validate RDATA against RRset's
class
-------------------------------------+-------------------------------------
            Reporter:  muks          |                        Owner:
                Type:  defect        |  UnAssigned
            Priority:  medium        |                       Status:  new
           Component:  libdns++      |                    Milestone:  New
            Keywords:                |  Tasks
           Sensitive:  0             |                   Resolution:
         Sub-Project:  DNS           |                 CVSS Scoring:
Estimated Difficulty:  0             |              Defect Severity:  N/A
         Total Hours:  0             |  Feature Depending on Ticket:
                                     |          Add Hours to Ticket:  0
                                     |                    Internal?:  0
-------------------------------------+-------------------------------------
Description changed by muks:

Old description:

> There are two variants of `RRset::addRRset()`:
>
> * `BasicRRset::addRRset(const Rdata&)`
> * `BasicRRset::addRRset(ConstRdataPtr&)`
>
> Both variants need to add the Rdata into an internal
> `std::vector<ConstRdataPtr>` collection.
>
> In the case of the first variant, it has to create another `Rdata`
> instance with a `ConstRdataPtr` managing it. For the second one, it
> simply pushes the passed `ConstRdataPtr` into the collection.
>
> When RDATA of a type mismatching its RR class is added to the `RRset`
> using the second interface above, it will add it without any validation.
> The first will fail with a `bad_cast` when it tries to create a `Rdata`
> copy.
>
> An example is adding RDATA to an RRset with `class=NONE` (such as
> required by UPDATE). The first method throws. The second method passes.
> This needs to be made consistent or at least documented.
> Apart from handing `class=NONE` specially, we should reject RDATA that
> mismatch the class/type.

New description:

 There are two variants of `RRset::addRdata()`:

 * `BasicRRset::addRdata(const Rdata&)`
 * `BasicRRset::addRdata(ConstRdataPtr&)`

 Both variants need to add the Rdata into an internal
 `std::vector<ConstRdataPtr>` collection.

 In the case of the first variant, it has to create another `Rdata`
 instance with a `ConstRdataPtr` managing it. For the second one, it simply
 pushes the passed `ConstRdataPtr` into the collection.

 When RDATA of a type mismatching its RR class is added to the `RRset`
 using the second interface above, it will add it without any validation.
 The first will fail with a `bad_cast` when it tries to create a `Rdata`
 copy.

 An example is adding RDATA to an RRset with `class=NONE` (such as required
 by UPDATE). The first method throws. The second method passes. This needs
 to be made consistent or at least documented.
 Apart from handing `class=NONE` specially, we should reject RDATA that
 mismatch the class/type.

--

-- 
Ticket URL: <http://bind10.isc.org/ticket/3263#comment:1>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development