From 0ea5fc66924303d1bf73ba283a383e2aadee02f2 Mon Sep 17 00:00:00 2001 From: neodarz Date: Sat, 11 Aug 2018 20:21:34 +0200 Subject: Initial commit --- pipermail/nel/2001-June/000423.html | 73 +++++++++++++++++++++++++++++++++++++ 1 file changed, 73 insertions(+) create mode 100644 pipermail/nel/2001-June/000423.html (limited to 'pipermail/nel/2001-June/000423.html') diff --git a/pipermail/nel/2001-June/000423.html b/pipermail/nel/2001-June/000423.html new file mode 100644 index 00000000..aaddac09 --- /dev/null +++ b/pipermail/nel/2001-June/000423.html @@ -0,0 +1,73 @@ + + + + [Nel] Hello and a Question + + + + + + +

[Nel] Hello and a Question

+ Vianney Lecroart + lecroart@nevrax.com
+ Fri, 15 Jun 2001 19:35:16 +0200 +

+
+ +
> Anyway, my question is, why did you develop your own distributed
+> architecture? Why not adapt/extend/use an existing solution such as
+> CORBA or DCOM (urgh, I know, win32 only). CORBA has a fair amount of
+> support and it can be coaxed to have the characteristics you need. I am
+> not saying you should change, I was just wonder what led you to make
+> your own? It just seems that the overheads of designing a naming and
+> authentication service are something that CORBA would have done away with.
+
+As you point out, DCOM isn't portable so we ruled it out :)
+As for CORBA: Having worked a fair bit with it while at college, I concluded
+that it lacked some of the functionality that we wanted. For example, we can
+have several instances of the same service running simultaneously on our
+shard and the naming service can either return us a list of all of them or
+select the least loaded one (which CORBA cannot do). What's more, we found
+that CORBA was excessively heavy for our requirements, which only require a
+corner of the system. Our naming service only took a couple of days to
+develop and remains entirely open to expansion, adaptation or tuning to keep
+up with NeL's evolving needs.
+
+We looked in depth into the ACE and TAO library (Real-time CORBA with TAO
+(The ACE ORB)). The webpage seems down www.cs.wustl.edu/~schmidt/TAO.html,
+but the lib is absolutely huge... one hour at least of compilation. We
+didn't want to be dependent on such a large library just for a naming
+service and a communication protocol.
+
+Regards,
+Vianney
+
+
+
+
+ + +
+

+ -- cgit v1.2.1