RE: [xri] Performance of XRI DSig

[Sakimura Nat <n-sakimura@nri.co.jp>]

John,

That developper in the past has dealt with XML DSig in the past with Java 5,
coupled with Apache library, and it was 8 times slower than plain siganture.
So, that probably is the paforane loss due to the canonicalization.

In Java 6 and current library, this nonesense does not happen.

If the developers have an impression that XML DSig is very slow,
it may have been an implementation specific thing.
That's what I wanted to convey.

At the same time, some systems are still working on that kind of
environment and may be hard to change it anytime soon.
That is anther reality.

=nat

________________________________________
差出人: John Bradley [jbradley@mac.com]
送信日時: 2009年6月12日 0:28
宛先: Sakimura Nat
CC: XRI TC
件名: Re: [xri] Performance of XRI DSig

Nat is that 8x slower for just XML DSig with Java 5, or 8x slower for
both?

John B.
On 11-Jun-09, at 5:03 AM, Nat Sakimura wrote:

> Hi.
>
> To address one of my concern, I have asked my colleague to test the
> performance of the C14N.
>
> Test a: Calculate XML DSig for 10,000 times
> Test b: Calculate Signature over a string for 10,000 times.
>
> OS: WindowsXP
> CPU: Intel　Core2Quad Q9650 3.00GHz
> Memory: 4GByte
> Java：jdk1.6.0_12
> (Tested on Eclipse)
>
> Signature Algorithm: RSA + SHA1
>
> [Measurement Method]
> Repeat the test 10 times and take the average.
>
> a. Take an XML DSig (javax.xml.crypto.dsig.XMLSignature) over an XML
> (DOM Structure)
> 50.353sec/10000　= 0.00503 sec/sig
>
> b. Take an Signature over an XML String (java.security.Signature)
> 49.300sec/10000　=　0.00493secs/sig
>
> So, on Java 6, impact of canonicalization is fairly small.
>
> That's a good news.
>
> Bad news is that the guy who did this testing told me that
> when he used Java 5 with Apache library, the performance was
> eight times slower. If that is true, it would be an issue.
>
> Best,
>
> =nat
>
>
>
>
>
> Result:
>
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe from this mail list, you must leave the OASIS TC that
> generates this mail.  Follow this link to all your TCs in OASIS at:
> https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php