[getdns-api] Priming the cache

William Chan (陈智昌) willchan
Wed Mar 13 16:38:42 CET 2013


On Wed, Mar 13, 2013 at 11:11 AM, Paul Hoffman <paul.hoffman at vpnc.org>wrote:

> On Mar 13, 2013, at 11:07 AM, William Chan (???) <willchan at chromium.org>
> wrote:
>
> > I think the feature is reasonable, but I don't understand why it should
> be part of the API. Can someone explain why this is the best place to put
> it? Why not keep this implementation specific?
> >
> > The reason I ask is, does this require all implementations of this API
> to implement a cache? That seems like unnecessary burden on
> implementations, since I can imagine many implementations simply proxying
> DNS queries to some other service that may implement caching.
>
> I was intending of adding the context function to note that not all
> implementations would have a cache. If you said "prime the cache with X"
> and the implementation had no cache, the call would return a new error,
> GETDNS_RETURN_NO_CACHE_TO_PRIME.
>

So we could do that. If implementing caching is optional, then do we get
much advantage from standardizing an API to modify the cache? I don't
really understand what value this return value provides. Presumably this
error is always returned given a certain implementation. That means that at
the point that I select my implementation, I need to know how to handle
this case, not dynamically during application runtime.

I think caching is very implementation specific. How large is the cache?
What's the eviction policy? What happens when the application tries to
preload too much into the cache? Etc etc. AFAICT, standardizing such an API
would either fail to make application developers agnostic of the
implementation, or standardizing this API would overly constrain
implementations and perhaps hurt performance.

So yes, we could definitely add this to the standard API. Do people think
there's much value though? I think it's a nice feature for implementations
to provide, but I am less convinced that we need to provide a standard API.
I think it'd be cool if someone explained why this is good to standardize
in the API. Just my two cents, I don't really care :)


> --Paul Hoffman
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.vpnc.org/pipermail/getdns-api/attachments/20130313/21b2189c/attachment.html>



More information about the spec mailing list