I don't think its a question of "unusable" but certainly things like a VPN
are one way to add some security around physical connection, but lots of
other scenarios require a bit more, or different, security.  e.g. if you
have 200 businesses connecting then the odds are that you won't be able to
agree on one VPN solution and want to see identity more directly through the
applications.

The point really is that well engineered systems have "ilities" and it is
the engineering that matters, picking one software implementation approach
or another will help (or hinder) in those decisions but ultimately there is
no magic wand that means "use X and your system will meet all its ilities"

On 04/06/07, Hitoshi Ozawa <[EMAIL PROTECTED]> wrote:

  +1
Makes me remember a comment made by a customer some years back
when he told me that he insists on using TR4 (token ring) over 1GB CDMA/CD
because token ring was more "reliable".

I, also, remember reading an article few years back on cache. Everybody
wanted to
put cache in their part of a product - e.g. chips, OS, and applications.
The article
debated on whether all these caches actually made the overall
application to the enduser
faster.
I think security is like a cache. If my application servers are all
going to be connected
only by a private network segment or by a VPN, do I need another level of
security? I do understand that there are some situations where there is
no underlying
security layers, but I was wondering how many such actual situations
exists to
warrant a technology as being labeled unusable.

H.Ozawa


Mark Baker wrote:
> On 6/2/07, Steve Jones <[EMAIL PROTECTED]<jones.steveg%40gmail.com>>
wrote:
>
>> To say that something that works over HTTP will exhibit "reliability"
>> sounds a little bit odd, given that HTTP is by design an unreliable
>> protocol.
>>
>
> REST concerns itself with properties of an architecture as a whole,
> not of individual features of the system. It has been understood for
> decades, that it is possible to build reliable systems with unreliable
> messages.
>
> Mark.
>
>

Reply via email to