Softpanorama (slightly skeptical) Open Source Software Educational Society

May the source be with you, but remember the KISS principle ;-)

The cost of patching

There are two sub-dimensions of this factor:

• Windows and linux are, and will remain for the foreseeable future, two most favorite OSes for attacks.

• Quality of recommended patch clusters produced by Sun is one of the top in the industry

Windows and linux are and will remain for the foreseeable future two most favorite OSes for attacks

In the large enterprise environment more frequent patching can double the maintenance  costs of the OS wiping any potential saving more effectively then anything else.  Cost of patching is a dirty secret of the Windows world and generally the requirement of patching the servers more frequently then once a quarter represents a very heavy burden (and huge costs overhead) for any large enterprise. That's why some of them are migrating parts of Windows infrastructure to Unix (including linux) despite high quality of recent Microsoft Server offerings (Windows Server 2003),  rich development environment that Windows provides, an excellent patch infrastructure and low cost of hardware. That also somewhat limits the area of linux deployment to front end Web server parks. The problem of patching of linux servers that are running complex enterprise applications like Oracle of SAP/R3 should not be underestimated. While often it is unclear to what extent the threats that are published and then hyped are real but still there is a significant level of internal paranoia in large enterprises that forces IT to react of them promptly. I would even suggest that professional deeply-technical evaluation of applicability of threats might be a useful cost saving measure, especially for enterprises that widely deploy Windows or linux. As critical applications can be affected patching necessitates creation of  rather expensive "quality control" infrastructure including servers specifically assigned for the testing of influence of particular patches on applications. Paradoxically due to relative security of UltraSparc architecture and availability of free recommended patch clusters Solaris on UltraSparc was a cheaper OS then linux in internet infrastructure roles. All those waves of security paranoia that rocked periodically Intel-based server world after each successful worm were almost unnoticed in UltraSparc space (as well in other RISC-based Unixes).  Actually many enterprise customers were able to run non-essential Solaris servers with only hardware support contracts or using cheaper third party support contracts and quarterly patching conveniently ignoring multiple security advisories from Windows and linux world without any detrimental effect and saving substantial money on IT security personnel.  With Solaris 10 free recommended patch clusters might be a thing of the past as getting them requires a support contract but quarterly patching schedule still is adequate for most servers. And security patches remain free. 

Problems with Red Hat kernels periodically are making headlines in major industry publications. For example in  May 26, 2006  article Red Hat Plugs Multiple Linux Kernel Flaws was published in eWeek:

The company is distributing updated kernel packages meant to fix 16 individual flaws present in the version 4.0 releases of its Red Hat Desktop and Red Hat Enterprise Linux OS software.

The company advised that all Enterprise Linux 4 users should upgrade their kernels to protect themselves from the security issues, 10 of which the Red Hat Security Response Team rated as "important," and six of which it tabbed as "moderate."

If compromised, the flaws could impact basic functions of the software, according to the Linux vendor.

Among the more serious issues reported by Red Hat were flaws in the software's IPv6 (Internet Protocol Version 6) implementation that could allow a local user to launch denial-of-service attacks on machines running the affected products.

Other important security included flaws in the software's ATM (Asynchronous Transfer Mode) module, NFS (Network File System) client implementation, and a difference in the "sysretq" operation of the OS with certain microprocessors, all of which could lead to different types of denial-of-service exploits.

As far as I cal tell monthly updates are the way of life for owners of Red Hat distribution. The latest critical update that I found was Aug 23, 2006 (see Red Hat update for kernel - Advisories - Secunia).
 

Quality of recommended patch clusters produced by Sun is one of the top in the industry

The key problem with patches is that they can affect other applications. Many large enterprises apply quarterly Sun patches for many years and do not experience any significant downtime. That is not true for many other vendors including major Sun competitors. IBM is noticeably worse the Sun (as in Saturday's evening phone call "Listen, automounter stopped working after AIX patches application on our boxes. What we should do now ?" ;-) . In turn Red Hat is noticeably worse then IBM AIX as for quality of patches and because it requires approximately three times more frequent patching for security reasons. For HP-UX quarterly application of recommended patch clusters is difficult as HP does not release patch clusters that often.  And even when it releases parches they are seldom applied.  So far "security via obscurity" worked well for HP-UX administrators and the standard industry practice for HP-UX boxes is to be  patched one a year if at all: any reasonable hacker has probably so intense dislike for HP-UX that they never port exploits to it :-).
 

Created Jan 2, 2005.  Last modified: February 28, 2008