Please don't give Sun's (ONC) RPC anywhere near that much credit. ONC RPC really pales in comparison to DCE RPC. First of all, ONC RPC uses XDR for serialization, whereas DCE RPC uses NDR. While, I would _love_ to see an RPC protocol based upon ASN.1 take off, at least NDR is multicanonical. Also, don't forget that the DCE RPC endpoint mapper integrates considerably more code to transparently handle things like authentication, authorization and directory services integration. The only thing ONC RPC brings to the table is GSS-API authentication.
What possible collection of technologies supercedes DCE/DFS??? Attacking TOG ignores the root of the issue. What other RPC service has integrated Kerberos authentication, authorization, and service registration in directory services for load balancing? All I see are a bunch of half-baked, buzzword-compliant piles of steaming horse manure that totally ignore security (e.g. SOAP). RPCSEC-GSS brings in K5 authentication, but none of the advanced directory, authorization and auditing services integration that DCE provides.
And what, pray tell, could possibly supercede DFS??? There is absolutely nothing comparable to DFS. Certain big companies will claim they have replacements, but only AFS and DFS provide the administrative unit of a fileset/volume, which is a critical unit of abstraction when performing management operations on distributed filesystems scaling into the trillions of files. Sure, NFSv4 has an atrocious hack for mountpoint redirection, but without a VLDB you don't gain any of the HA features, nor any of the flexibility. And, only DFS provides things like multiple read/write replicas of filesets, distributed file locking, and sync on write semantics. SAN filesystems, and distributed block-level raid filesystems, such as IBM GPFS and Sun ZFS, address completely orthogonal problems to DFS. I'm sick of hearing people pushing such solutions as DFS replacements. They are perfect for high-throughput access to scientific and media-oriented applications, but they will never provide the same levels of management, or availability as a truly distributed filesystem like DFS.
Frankly, the argument that technology should be ignored because it is old has proven its ridiculousness way too many times over the years. Far too many very interesting and worthwhile technologies are being abandoned by executives who have no appreciation for their technological value. HP's track record in this area is especially abysmal.
Slashdot Mirror
Please don't give Sun's (ONC) RPC anywhere near that much credit. ONC RPC really pales in comparison to DCE RPC. First of all, ONC RPC uses XDR for serialization, whereas DCE RPC uses NDR. While, I would _love_ to see an RPC protocol based upon ASN.1 take off, at least NDR is multicanonical. Also, don't forget that the DCE RPC endpoint mapper integrates considerably more code to transparently handle things like authentication, authorization and directory services integration. The only thing ONC RPC brings to the table is GSS-API authentication.
What possible collection of technologies supercedes DCE/DFS??? Attacking TOG ignores the root of the issue. What other RPC service has integrated Kerberos authentication, authorization, and service registration in directory services for load balancing? All I see are a bunch of half-baked, buzzword-compliant piles of steaming horse manure that totally ignore security (e.g. SOAP). RPCSEC-GSS brings in K5 authentication, but none of the advanced directory, authorization and auditing services integration that DCE provides.
And what, pray tell, could possibly supercede DFS??? There is absolutely nothing comparable to DFS. Certain big companies will claim they have replacements, but only AFS and DFS provide the administrative unit of a fileset/volume, which is a critical unit of abstraction when performing management operations on distributed filesystems scaling into the trillions of files. Sure, NFSv4 has an atrocious hack for mountpoint redirection, but without a VLDB you don't gain any of the HA features, nor any of the flexibility. And, only DFS provides things like multiple read/write replicas of filesets, distributed file locking, and sync on write semantics. SAN filesystems, and distributed block-level raid filesystems, such as IBM GPFS and Sun ZFS, address completely orthogonal problems to DFS. I'm sick of hearing people pushing such solutions as DFS replacements. They are perfect for high-throughput access to scientific and media-oriented applications, but they will never provide the same levels of management, or availability as a truly distributed filesystem like DFS.
Frankly, the argument that technology should be ignored because it is old has proven its ridiculousness way too many times over the years. Far too many very interesting and worthwhile technologies are being abandoned by executives who have no appreciation for their technological value. HP's track record in this area is especially abysmal.