On 02/08/16 21:23, ellisonanne wrote:
> Github user ellisonanne commented on a diff in the pull request:
> 
> https://github.com/apache/incubator-pirk/pull/43#discussion_r73229948
>
>  --- Diff:
> src/main/java/org/apache/pirk/querier/wideskies/QuerierDriverCLI.java
> --- @@ -55,7 +58,7 @@ public static final String PAILLIERBITSIZE =
> "paillierBitSize"; public static final String BITSET = "bitSet"; 
> public static final String CERTAINTY = "certainty"; -  public static
> final String QUERYNAME = "queryName"; +  public static final String
> QUERYID = "queryID"; --- End diff --
> 
> See previous comment on the QuerierDriver:
> 
> The queryName corresponds to the schemaName in the query-schema.xsd
> file not to the user-given identifier for the query -- the identifier
> for this query is actually the queryNum. Notice that queryNum is
> currently a double -- if we want to start appending dates or whatever
> else to it, we need to change its type to String throughout.

I had been wondering why this was a double -- seems strange.

I was considering changing it to be a UUID, what do you think?

While it may be ok given Pirk's policy on trust to use an RFC 4122
type 1 UUID, which would reveal to the responder the machine IP address
and timestamp in the query, that would make it more difficult for higher
level applications to preserve the anonymity of a querier, so I'm
inclined to opt for the slightly more time consuming type 4 (random) UUID.

I think the performance difference will be lost in the noise.

Thoughts on this too?

Regards,
Tim

Reply via email to