IE 8 Beta 2 Impressions

Technorati tags: IE8, beta, web browsers

Internet Explorer 8 beta 2 came out just before Google Chrome, and both have several features in common.  However, Chrome is still built on a known-exploitable code base that even Apple had already patched in Safari, so I can’t take that seriously as a safe edge-facing product.  I also don’t like vendors who silently push updates without user control, who who are slow to tell you what their new code does.

Features

Beta 2 feels a lot more new-featured than beta 1, partly because it is, and partly because the new UI is catching up with what is inside the code.  There’s enough difference and improvement over IE 7 to be worth a look, even if you aren’t interested in the beta testing and feedback process – but it is in beta, so take care if testing it your “real” computer.

I enjoyed the safer address bar URL display in beta 1, but there’s more URL-handling goodness in beta 2, especially when it comes to revisiting where you’ve been, keeping track of which tabs were opened from where, and so on.  There’s also the InPrivate mode if you don’t want your history retained.  And I love the new search entry area’s drop-down set of search engine icons, so I can change engine on the fly for each search instead of having to set and change this.

A subtle but significant joy is improvement of Standards vs. IE 7 modes.  In beta 1, one had to exit the entire browser to “change gears”, which seemed odd as each tab runs in its own process, including a mix of tabs in Protected and non-Protected modes.  In beta 2, you can not only mix Standards and IE 7 modes across a tab set, but have the browser remember and automatically apply the correct mode for a given site or domain.

It’s a bit strange to still have no “advanced” download support, i.e. the same ability to resume interrupted downloads that we had in the 386 era’s BBSs transfer protocols.  After all, ftp does as last embrace such things, as well as the ability to download from multiple points in the same file for speed.  Why should we still need tacky and often spyware-riddled add-on tools to do this?

Stability

On stability, it’s a bit of a mixed bag, compared to beta 1.  It now handles the Trend SysClean engine download link a bit better; it still shows the Windows executable code content within the misnamed .COM file as a web page, but no longer crashes and can Save As correctly.  Some other sites that didn’t work in beta 1, such as on-page UI buttons to send comments from some blogs, now work correctly in beta 2.

On the downside, IE still struggles to close when there are many tabs open (typically I have 20-30 open per window at a time), and can fall into a doomed loop of retrying the same page that crashed it.  I’ve also seen the whole browser become terminally unresponsive, and not always when trying to close down multiple tabs.  That can cause a storm of crash dialogs etc. if it tries to recover all the tabs that were open.

When the browser becomes unresponsive, I either see System and IExplore pegged at 50% CPU each, or (more often) > 90% of CPU time correctly in the idle loop, while IE still behaves as if it’s being hogged.

All of this testing has been on XP SP3 with IE8 running in Standards Mode, and set to prompt on active content.  That can lead to multiple tabs stalling on such prompts at the same time, which may bulge resource use or lead to some sort of race condition.  I suppress 3rd-party enhancements via the IE 6 era Tools, Options, Advanced setting, so there shouldn’t be any BHOs or toolbar code.

I’m certainly enjoying beta 2, but wonder whether there will be more builds to test between now and RTM.  IMO, there need to be, if only to pin down and fix the stability issues that remain.

IE Needs True "Blank" Page

Technorati tags: IE8, beta, wish, Microsoft, web browser

In a malware-aware 21st century, IE really needs a true “NO home page” option, and not load an infectable .HTM file from a known filespec.

After all, if one wants to load a local .HTM as “home page”, perhaps to effect some scripting at the start of a browser session, then one can point to such a file as “home page”.  That is far clearer (”principle of least surprise”) than creating the impression of a “blank” home page that is not certain to be “blank”.

In other words, the “blank home page” option in IE8 should show nothing, and not load any file as the “home page”.  If IE8’s code requires something to chew on, then IE8 could spawn a fresh blank .HTM of random name from boilerplate code, and load that.

Please not let’s hear a lot of “mitigation noise” such as “an attacker would have to…” or “blah blah permissions” this, or “blah blah reduced user rights” that.  Non-existence trumps “security”!

IE8: Summary and Launch Pad

Technorati tags: IE8, beta, wish, Microsoft, web browser

I’ve been testing Internet Explorer 8 beta 1, and so far it looks like IE7, but with more goodly oomph under the hood.  It’s a relatively early beta, so I expect the look to change; at present, the UI is a mix of standard-looking IE7 with very new-look dialogs here and there,

This post will be retro-edited to serve as a launch point for bugs and suggestions for IE8, much as I did for Vista here.

Name This Blog

Technorati tags: Blogging

At the risk of sounding like a self-important vendor spending millions on announcing a changed product name, this post is a muse about where to take this blog.

The “problem” is that this blog hosting service has the feature set I want to use, and I want to do more than write about Linux.

I like the symmetry of “blog“, Vista Curve” and “Linux Curve“, and it’s not obvious as to how this third blog will be differentiated with the others.  I was thinking of calling it “The Long View”, to include ideas for Windows 7, IE 8 (before that gets feature-frozen, if it hasn’t already), Linux, and the still-neglected area of maintenance OSs. 

But I also want to use the web site structure offered by WordPress to cover timeless “how do I…” issues definitively, e.g. how to defend against USB infectors.  The things I’d have done on my old web sites in the old days, and as these are not particularly forward-looking, they don’t fit “The Long View”.

Perhaps it depends on how well WordPress can offer alternate faces for web and blog content.  If it’s good at that, I’ll use that to separate the gritty reality of the site from the more futuristic blog.  Or maybe I can shell the “Linux Curve” within the “Long View” blog, much as my old Win9x site was shelled under a new parent site as I turned attention towards XP.

You may think it’s odd that an MVP has a blog dedicated to Linux, but IMO, any “long view” (5+ years) has to count in alternatives to Windows, and to PCs for that matter.  Microsoft’s research labs would be doing the same, and they aren’t just cooking up new versions of MS Office and Windows.  But as users, we don’t have the same ability to design our futures – we’d have to choose from off-the-peg solutions that could be so different that today’s judgement of acceptability cannot be assumed to hold.

Editing menu.lst Revisited

Technorati tags: Linux, grub, Ubuntu

In response to this entry, Cenk Kulacoglu asks why I spawned a new CLI session rather than just run a terminal instance from the appropriate menu.  The reason is simple; I didn’t think of that  :-)

In addition, one doesn’t have to use the vi text editor.  In fact, you can at least initiate most of this via the GUI’s menus, though the lack of an equivalent to Vista’s right-click, “run as admin” makes you more CLI-bound than should be necessary.

The easier way I now use is as follows; start a command line window from the appropriate menu, navigate to /boot/grub, sudo to chmod the menu.lst file for editing, then sudo gedit menu.lst to edit the file in a more familiar style of text editor.  After saving the edits, chmod the menu.lst back to protect it, and you’re done.

The stumbling block was having to “smell” that the nice text editor seen via the GUI menu, is named gedit, so that one can invoke it from the CLI.  It’s necessary to invoke it from the CLI rather than the GUI menu, because the GUI menu lacks a “run as admin” (i.e. with root access) facility.

Linux File Attributes

Technorati tags: Linux

The chmod command operates on the Linux equivalent to “attributes” in DOS, but the syntax differs from the DOS attrib command.

These attributes (or “modes”?) are conceptually similar, being a bitmapped value with each bit specifying an attribute of the file.  When you do an ls -g to see more details for a directory listing, they are displayed something like this:

-rwxrwxrwx

Each group of “rwx” corresponds to a single octal digit, i.e. values 4 + 2 + 1.  If the bit is set, the flag is shown; if it is zero, then – is shown instead.  So rwx = 7 and means reading, writing and execution permissions are allowed, whereas r– mean only reading is allowed. 

These permission bits are then repeated for different contexts; from left to right, these are for owner, group and user, according to this syntax reference.

I see there is support for syntax closer to the “Attrib -r -s -h” familiar to DOS users, but I used octal values when I needed to chmod something, e.g…

chmod 0777 menu.lst

chmod 0644 menu.lst

…that result in…

-rwxrwxrwx

-rw-r–r–

…respectively.

Ubuntu: Editing grub’s menu.lst

Technorati tags: Linux, grub, Ubuntu

If you’ve just set up a dual-boot between Windows and Ubuntu (in this case, 7.04) to try out the latter, then you will probably want to edit the settings file that controls the grub system-level boot manager.  You may want Windows to boot by default, and more menu time to override that default.

As seen from Ubuntu, the file you are after is in the /boot/grub directory, and is called menu.lst (where “l” is lower-case L, not the digit 1).  But because this is outside your home subtree, you will need root privileges to edit it. 

I didn’t find a way to do this in the GUI (e.g. an equivalent to Vista’s right-click, “Run as administrator”) so I pressed Ctl+Alt+F1 to get to a bash shell (a Linux equivalent to Command.com or Cmd.exe).  From here it is easy; precede any command needing root access with the sudo command, and after passing a root password prompt, that works.

After logging in to my user account, I entered the following commands to get to where the file was, list it to see what permissions were in effect, then change these to facilitate my changes:

cd /boot/grub

ls -g

sudo chmod 0777 menu.lst

ls -g

This took the safety catch off menu.lst so I could save changes to it.  Then I used vi to edit the file, changing the order of boot entries so that Windows came first (I could have left it where it was and specified it by number as the new default), and then changing the timeout statement to “timeout sec    20″ for 20 seconds (the word “sec” is required). 

I’ll leave the gory details of editing in vi as an exercise for the reader, but basically it’s a to add text, Esc to go back to command mode, and :wq to save and exit.

Thereafter I restored safer attribute flags via this command…

sudo chmod 0644 menu.lst

…pressed Ctl+Alt+F7 to switch back to the GUI, and was done.  Whew!

Ubuntu 7, XP and Intel G33

Technorati tags: Linux, XP, G33, dual boot

I’ve just done a dual-boot build with XP Home SP2 and Ubuntu 7.04 Feisty Fawn on a PC based on Intel’s Frostburg motherboard.  This is an ATX PCI Express DDR2 motherboard, based on the G33 chipset, which in turn includes the ICH9 IO controller. 

Intel G33 and ICH9

The ICH9 is significant, because XP’s \i386 has no native support for it – so a Bart mOS CDR based on the XP SP2 code set cannot see S-ATA hard drives connected to it, and the XP SP2 installation process requires the “press F6 and provide drivers on diskette” step else it too will find no S-ATA hard drives on which it can install (it does however see external hard drives attached via USB!).

This creates a maintenance quandary for Windows installations on this chipset, especially as this has been the motherboard I have used for all new PCs since it became available here a few months ago.  Bart will work if I drop the hard drive in an older and compatible motherboard.  WinPE 2.0 works, but lacks RunScanner functionality.  New Linuxen can write NTFS, which makes it close enough to WinPE 2.0 to use as a mOS; the downside is no registry support at all, but the upside is that it may be easier to set up as a software platform than WinPE, and what I learn is directly applicable to its use as a production OS.

When I started looking at Linux again a few months ago, I threw a motley pile of distros of various vintages at the G33, and none of them stuck (none booted “live”, none could install), presumably because they were made before the G33.  The exception was Ubuntu 7.04, which booted and installed, but the hard drive installation wouldn’t boot.

Vista

I build Vista using the WAIK tool set; I built a reference installation, avoid activating it, initialize via Sysprep, then harvest that via ImageX from WinPE.  New PCs are then built by applying the image via ImageX from WinPE, after partitioning via BING (which I don’t install) and formatting C: to NTFS from WinPE (the Convert tool appears not to work).

Originally I built this as a Vista PC to replace an aging Win98SE PC that was used to run an old version of Flexisign.  The client is a small mom-and-pop signwriting shop that is built around a pair of Summa vinyl cutters that were bought second-hand, along with the Flexisign software that they use.  They can’t afford to buy this software anew; even the upgrade is too costly.

The old version needs a parallel port dongle and works only on the old Win98SE PC and within a Win98SE VM on an XP PC, but the new Frostburg motherboard has no parallel port hardware. The newer version runs fine in XP and Win98SE, but not in Vista, so the initial solution was another Win98SE VM within Vista to run it that way.

After some time on site, I had peer-to-peer networking in effect for both VMs and physical PCs (the trick being how one applies sharing permissions in Vista).  I went home happy, only to find a bunch of increasingly desperate email messages saying this and that doesn’t work, the network file sharing wasn’t working, and “please can we go back to XP”.  Bye-bye Vista; it doesn’t matter how pretty you are, if you can’t host applications and do the basic bread-and-butter stuff, you aren’t a solution here.

Ubuntu 7.04

Given the long-term implications of using vital software that is already old and too costly to replace, we decided to build a dual-boot XP and Linux system.  The clients don’t really like Flexisign that much, and are happy to use something else if it is legally free; that’s more likely to happen on Linux than Windows.

I started by using BING to image the Vista 32G C: to logical volume E: so that I could undo if things got even worse.  Then I deleted the primary C: partition, created a new 20G primary for XP, slid that forward to reduce head travel to the extended partition, and left 12G free space for the Ubuntu installation.

Before installing Ubuntu, I copied off all files from logical volumes D:, E: and F: in case the Ubuntu installation process botched those for some reason.  The last time I installed Ubuntu back in version 5.xx, the installer was quite cavalier, defaulting to wiping the entire hard drive, and I’d recently done data recovery for another client who had botched additional 300G and 500G hard drives when installing Linux.  So I took care to maintain undoability!

The installation process was very easy and went flawlessly, with no additional drivers required.  This time there seemed no risk of botching existing hard drive contents, though I was careful to go “custom” and watch all settings and checkboxes along the way.  I created an 11G primary for Linux and 900M swap partition, which thus filled up the partition table.

If you read my 10 Linux Toe-Stubbers article, you’ll know I anticipated some hard work ahead, if I wanted to “see” the same logical volumes as used from the Windows installation.  But that was easy; Ubuntu shows all of these volumes, already mounted, as icons on the desktop!

I had to roll up my sleeves to edit the grub menu.lst file from Ubuntu, so that it would default to Windows XP (which I hadn’t installed yet) and pause long enough to select Ubuntu instead.  I could see the required file, but it wasn’t obvious how to access it with the required root privileges so that I could save my changes to it.

So I pressed Ctl+Alt+F1 for a bash shell, and from there I figured out how to log in as root and then use the vi text editor to edit the file.  With my Linux Bible on my knees opened at the relevant pages, I trudged through vi’s terminal-era keystrokes to make my changes, then saved and exited.  The menu now works and defaults correctly, but the timeout value of 15 does not seem to give 15 seconds of menu time – something I’ll fix later.

Everything works in Ubuntu except the integrated LAN adapter – so I’ll have to download possible vector graphics applications etc. from the XP side of the fence.  But for now, I’m happy enough to move on.

XP Home

In contrast, the installation of XP Home was a trickier affair.  Before I discovered WAIK, I used to build XP systems interactively; now I do so by applying a reference build via WAIK’s tools, much as described for Vista.  Normally I use a 7.9G FAT32 C: for XP, but they needed more C: space (thanks to Garmin and Google Earth bloat factors), so I formatted the 20G C: to NTFS in order to stay with paging-friendly 4k clusters.  Then I applied my XP installation image.

This resulted in an installation that booted into a BSoD STOP error, including from Safe Cmd; also, the Recovery Console and DOS Mode Boot.ini menu entries didn’t work.  None of this surprised me; after all, the motherboard chipset was different and beyond the native capabilities of the \i386 file set, and as I’d changed the partition size, any PBR images used to boot Recovery Console or DOS Mode would be doomed. 

DOS Mode won’t run from NTFS anyway, so the plan would be to delete that option from Boot.ini, and renew the Recovery Console by re-installing that from the generic OEM XP CD that I’d bought for the build.

But first, I had to do a Repair install to create a code driver code set that matched the newer motherboard.  That in turn failed, as the base code set could not see the S-ATA hard drive through ICH9.  So it was off to Intel’s web site to download the appropriate drivers, and make a 1.44M diskette to wave at the F6 opportunity.

With a diskette drive hooked up to host the diskette, I retried the repair install, and this time, all went well.  The F6 process appears to ignore the diskette at first, but later stops and lists the drivers found on it, at which point you need to know which option to choose after pressing S to add such things (I chose the first entry, i.e. ICH9R).

After that, all went as well as can be expected for an XP repair install; all patches and IE 7 lost, dialogs to carefully step through so as to re-assert regional settings and skip the creation of unwanted new user accounts, then fixing of lost settings such as “don’t automatically restart on errors”, “don’t do System Restore on logical volumes D:, E:, and F:”, etc.

A pleasant surprise was that the XP imaging and repair install had not botched the grub system-level partition manager!  Not only had that survived, but my re-use of the sample Windows boot entry settings actually matched the installation I created, so that worked as well.

The activation process was sticky, too; I could not log into XP without activating first, and activation over the Internet failed because the installation didn’t have drivers for the LAN adapter as yet – a classic “don’t kick away the ladder” kind of meta-bug.  If I didn’t know to select the “activate by telephone” option, I’d have been stuck at this point, but that process went smoothly.

Overall impressions

I was impressed with Ubuntu 7.04; it not only took to the new hardware with aplomb, but anticipated and catered for my desire to access to other hard drive volumes without me having to manually mount them, and then bang my head against automating this process on every Ubuntu boot.

It may seem unfair to expect XP SP2 Home to work as well with hardware that didn’t exist when the code base was created, but this CD appears to have been re-made since the original SP2 (the activation input key is longer, probably a recent change as Microsoft ran out of product key space for the product).  It would have been nice, though at the cost of testing etc., had core driver support for new chipsets been added – especially as sales of XP have been extended due to demand.

I don’t know how easy it will be to get Intel’s LAN adapter working in Ubuntu, or what other problems may have to be solved down the line, but it’s a good start!

My 10 First Linux Toe-Stubbers

Technorati tags: Linux, Windows to Linux

The first 10% of what you learn takes 90% of the effort, and as it can never be undone (merely “painted over”) it’s important to start right.

Some of the first 10 things I needed to learn about Linux are basics, others are not; i.e. less-basic things that have to be got right, before anything else can follow.

My perspective is of a Windows user wanting to dual-boot to Linux, with most of the hard drive’s contents being visible to both OSs.  As a builder of such systems, I have strong Windows skills, good general computing sense, and very few Linux smarts!

1  Where’s “Ctl+Alt+Del” ?

Since Windows 95 it’s been instinctive to do the “Vulcan grip” of Ctl+Alt+Del to bring up the task manager, so that errant tasks can be killed.  Almost as soon as I’d booted a Knoppix Live CDR, I needed the equivalent Linux skill, and it’s very frustrating without it.

As far as I know, the answer is various key combinations that will launch command interpreters (which are called “shells’ in Linux), and from there, I’d be expected to enter commands to list tasks to read the corresponding ID, then kill particular tasks via that ID.

2  Where are the “drive letters” ?

Since Windows 95, the DOS file system navigation of drive letters, directories and files has been abstracted to a namespace of folders and objects.

UNIX and Linux have always had that level of abstraction built into the file system, such that what appear to be non-file-system objects in the modern Windows namespace will appear to be syntax-valid entities within the Linux file system. 

We’ve had a whiff of that with DOS devices like LPT1:, CON:, etc. that were carried over from UNIX, but Linux takes that further; such devices can be “pathed” as if they were directories and files within the file system structure.

So not only does one use the “file name” hda1 to refer to the first partition on the first hard drive, it’s also more fully expressed as /dev/hda1 as within the dev “directory”.  Things that are not actually files or directories, have to be “mounted” as items within a directory, so that file system syntax can be applied.

3  How do I know what type a file is?

DOS had strong links between file types and file name extensions, but no system to extend this to automatic association of file types to arbitrary programs, as was added in Windows 95.  Alas, this invaluable safety feature is being eroded within modern Windows.

In contrast, UNIX and Linux don’t rely on visible file name extensions to associate files with programs that “open” them.  It may sometimes user these by convention, e.g. .gz or .tar for archives, but most of the time there is no obvious indication of what a file is, or what it can do.

4  How do I set default OS in boot loader?

Now we get into specific things that need to be done, whether they are “basic” or not. 

Windows has no generic system-level boot management tools; all of its boot selection mechanisms start within the OS boot and work only for peer and older Microsoft OSs.

Thus when you set up Linux and Windows as a dual-boot, it is Linux that generally brings the system-level boot manager to the party; either lilo or grub.  And in time-honored software vendor tradition, it will default to booting itself unless you choose otherwise.

So one of the first things a Windows user playing with Linux as alternate boot will want to do, is edit the boot loader’s configuration file to set Windows to boot first, as well as clean up what is often rather mucky menu text.

It seems as if no Linux disto anticipates this need; you may have plenty of friendly “getting started” help stuff, but do you think “edit boot manager settings” will appear as an icon or menu drop-down?  I never did find out how to that in Ubuntu 5.x, and just gave up in the end.

5  How do I stop Linux dialing up all the time?

Your mileage may vary on this one, but this was a major PITA in Ubuntu 5.xx – once set to be able to dial up to the Internet, it would do so on every boot. 

In a place where local phone calls to ISPs are billed per second by the local telco monopoly, this is a Bad Thing – especially as any unattended reboot will default to Linux and do this again.  There’s no “can I do this?” dialog to cancel, either; it would just go ahead and help itself to your phone bill, for no apparent reason.

I never did find an answer to this, other than never boot Ubuntu again.  I could have tried harder, but why should I have to?  This should be something intuitive to manage via a control panel of some kind. 

6  Doesn’t Linux know what an extended partition is?

To dual boot Windows and Linux on a single hard drive, with data volumes visible to each, one needs at least three of the four available partition table entries; one for Windows OS, one for Linux OS, and an extended partition containing the data volumes visible to both.

However, some Linux distros expect to hog three out of these four partitions for itself; one for the Linux system, one for swap space, and one for user data.

I know the “extended partition” is part of Microsoft OS space, rather than system space, but as Linux already understands and accesses FATxx and NTFS file systems, it’s surely not asking too much for it to be able to use logical volumes within an extended partition, rather than insisting of hogging 3/4 of the system partition table?

7  How do I mount arbitrary hard drive volumes?

When you first start up Linux via dual-boot, you won’t see your “Windows” partitions, until you mount these.

So finding out how to mount such volumes is one of the first things you need to do – but like (4), even the friendliest distros orientated towards converting Windows users do not seem to anticipate this need.

Good news: Ubuntu 7.04 automatically finds and mounts accessible hard drive volumes, and lists them as desktop icons for easy access!

8  How do I write and run “batch files” ?

I know you can enter shell commands into a text file and use that as a script, but the mechanics to do that are not obvious; where to save such files, what file name extension to use, and how to invoke it again.

This follows on from (7), i.e. once you figure out the command to mount your data volumes, you want to use a command that is easier to remember than what looks like modem line noise!

9  Can hard drive volumes be auto-mounted on boot?

One wants shared access to local data storage to be as transparent as possible, i.e. have such hard drive volumes mounted when Linux starts.

In contrast, Windows is too good at this; it automatically assumes every visible hard drive volume is part of the system, and starts groping content and writing to disk as soon as it starts up – which poses malware and data risks in various maintenance contexts, driving the need for a competent maintenance OS.

As noted in item 7, Ubuntu 7.04 solves this problem out of the box by default.

10  Where’s the “startup axis” ?

Windows has an absurdly large number of ways to integrate arbitrary code so that it runs automatically, and this is a bane of malware management.

But it’s not at all obvious where or what the equivalent(s) in Linux may be, i.e. how arbitrary commands, scripts or programs can be run whenever Linux starts up.

More Focus, Less History

Technorati tags: Blogging

Have you noticed how little of what you read, is about what you wanted to know about?  Firstly, there’s pointless “friendly” waffle like the two sentences you have just read.  Then there’s history of how things used to be, but aren’t anymore.  Then there’s background that may or may not be useful.

To some extent, one expects this on commercial sites, where a single article like this will span several pages that have to be accessed one after the other – with 50% screen area taken up by banner ads etc.  Blogging discourages that, as each post ideally fits on one page.

For example, many articles on quantum physics may start with Plato’s atoms and trace the history from there onwards.  Or an article on malware prevention may waste time on virus/worm/trojan taxonomies that are way past their sell-by date, the history of the first known virus, etc.  Or a book on “how to use Linux” will start off on Linux politics, which tells you nothing about how to use it.

I’m as prone to digression and waffle as anyone, but I’ll try to focus when I write technical articles here, and link out anything that isn’t directly about what the subject line claimed the article is about.