In article <10cmovf$3a740$1@dont-email.me>,
Arne Vajh|+j <arne@vajhoej.dk> wrote:
On 10/13/2025 10:03 PM, Lawrence DrCOOliveiro wrote:
On Mon, 13 Oct 2025 21:20:43 -0400, Arne Vajh|+j wrote:
On 10/13/2025 8:20 PM, Lawrence DrCOOliveiro wrote:
On Mon, 13 Oct 2025 19:26:56 -0400, Arne Vajh|+j wrote:
Enterprises with a need to document support can not just hire a
random consultant when the need arrive.
If something is mission-critical and core to their entire business,
they want a staff they can rely on, completely, to manage that
properly.
Few/no CIO's want to support the hundreds of millions of lines
of open source code their business rely on themselves.
The whole point of having all that code is that they didnrCOt need to write >>> it themselves.
Yes. But they want free beer more than free speech.
You have to take responsibility for your own business, donrCOt you?
They don't want to write or maintain their own OS.
They don't want to write or maintain their own platform
software (web/app servers, database servers, message queue
servers, cache servers etc.).
They don't want to write or maintain their own tools
(compilers, build tools, IDE's, source control, unit
test frameworks etc.).
None of that stuff is their business.
They want to focus on their business the applications
that help them produce and sell whatever products
or services.
Every single one of the FAANG companies do all of those things.
At Google, we used to joke that, "not only does Google reinventYou're kinda going in circles here by arguing that very big companies
the wheel, we vulcanize the rubber for the tires." Spanner, Piper/Fig/Jujutsu, Prodkernel/ChromeOS/Android, CitC, gunit, Go
(not to mention the work on LLVM/Clang), Blaze/Bazel/Skylark,
etc, are all examples of the things you mentioned above. And
that's not even to mention all the custom hardware.
For organizations working at hyperscale, there comes a point
where the off-the-shelf solutions simply cannot scale to meet
the load you're putting on them.
At that point, you have no choice but to do it yourself.
I believe Arne's point was the fairly obvious one that a retail
chain or a hospital chain does not need to and cannot afford to
maintain, for example, their own operating system.
On Wed, 15 Oct 2025 15:33:20 -0500, Craig A. Berry wrote:
I believe Arne's point was the fairly obvious one that a retail
chain or a hospital chain does not need to and cannot afford to
maintain, for example, their own operating system.
Do you think that is hard to do?
In article <10cmovf$3a740$1@dont-email.me>,Few companies are like Google.
Arne Vajh|+j <arne@vajhoej.dk> wrote:
On 10/13/2025 10:03 PM, Lawrence DrCOOliveiro wrote:
On Mon, 13 Oct 2025 21:20:43 -0400, Arne Vajh|+j wrote:
Few/no CIO's want to support the hundreds of millions of lines
of open source code their business rely on themselves.
The whole point of having all that code is that they didnrCOt need to write >>> it themselves.
Yes. But they want free beer more than free speech.
You have to take responsibility for your own business, donrCOt you?
They don't want to write or maintain their own OS.
They don't want to write or maintain their own platform
software (web/app servers, database servers, message queue
servers, cache servers etc.).
They don't want to write or maintain their own tools
(compilers, build tools, IDE's, source control, unit
test frameworks etc.).
None of that stuff is their business.
They want to focus on their business the applications
that help them produce and sell whatever products
or services.
Every single one of the FAANG companies do all of those things.
At Google, we used to joke that, "not only does Google reinvent
the wheel, we vulcanize the rubber for the tires." Spanner, Piper/Fig/Jujutsu, Prodkernel/ChromeOS/Android, CitC, gunit, Go
(not to mention the work on LLVM/Clang), Blaze/Bazel/Skylark,
etc, are all examples of the things you mentioned above. And
that's not even to mention all the custom hardware.
For organizations working at hyperscale, there comes a point
where the off-the-shelf solutions simply cannot scale to meet
the load you're putting on them.
At that point, you have no choice but to do it yourself.
In article <memo.20251014170713.10624x@jgd.cix.co.uk>,
John Dallman <jgd@cix.co.uk> wrote:
In article <10ckadi$7dr$1@reader2.panix.com>,
cross@spitfire.i.gajendra.net (Dan Cross) wrote:
In article <memo.20251011151314.10624m@jgd.cix.co.uk>,
John Dallman <jgd@cix.co.uk> wrote:
Our stuff does gain significantly from 64-bit addressing; I could
believe fields that didn't need 64-bit gave up on Sun earlier.
I can see that. Personally, I really liked Tru64 nee DEC Unix
nee OSF/1 AXP on Alpha. OSF/1 felt like it was a much better
system overall if one had to go if swimming in Unix waters,
while Solaris felt underbaked.
I was happy with it, but a very experienced Unix chap of my acquaintance
reckoned "It doesn't run - it just lurches!" regarding it as a
Frankenstein job of parts stitched together.
Ha! I can sort of see why they'd say that. It definitely had
odd bits of Mach and System V seemingly bolted onto it. Overall
though I thought it was a good system.
To bring it back to VMS (and sheepishly admit a good bunch of
the recent drift is my own) We had an Alpha running OpenVMS AXP
1.2, or whatever one of the earlier versions was;
On 10/15/2025 8:16 AM, Dan Cross wrote:
In article <10cmovf$3a740$1@dont-email.me>,
Arne Vajh|+j <arne@vajhoej.dk> wrote:
On 10/13/2025 10:03 PM, Lawrence DrCOOliveiro wrote:
On Mon, 13 Oct 2025 21:20:43 -0400, Arne Vajh|+j wrote:
Few/no CIO's want to support the hundreds of millions of lines
of open source code their business rely on themselves.
The whole point of having all that code is that they didnrCOt need to write
it themselves.
Yes. But they want free beer more than free speech.
You have to take responsibility for your own business, donrCOt you?
They don't want to write or maintain their own OS.
They don't want to write or maintain their own platform
software (web/app servers, database servers, message queue
servers, cache servers etc.).
They don't want to write or maintain their own tools
(compilers, build tools, IDE's, source control, unit
test frameworks etc.).
None of that stuff is their business.
They want to focus on their business the applications
that help them produce and sell whatever products
or services.
Every single one of the FAANG companies do all of those things.
At Google, we used to joke that, "not only does Google reinvent
the wheel, we vulcanize the rubber for the tires." Spanner,
Piper/Fig/Jujutsu, Prodkernel/ChromeOS/Android, CitC, gunit, Go
(not to mention the work on LLVM/Clang), Blaze/Bazel/Skylark,
etc, are all examples of the things you mentioned above. And
that's not even to mention all the custom hardware.
For organizations working at hyperscale, there comes a point
where the off-the-shelf solutions simply cannot scale to meet
the load you're putting on them.
At that point, you have no choice but to do it yourself.
Few companies are like Google.
For a few reasons:
[snip]
3) Google is not just a company using IT to produce
products/services - Google is also a company doing
IT for other.
Google Search is an IT user where it is not a given
that they want their own distro.
But Android and ChromeOS is Google delivering an
OS to other. The OS is their business in that case.
And one facet of GCP is that Google is taking
over OS support from Redhat/Canonical/SUSE when
companies moves their workload from on-prem to
GCP managed services. Linux support is their
business.
My napkin calculation / RNG says you will need more than a million
Linux instances for the math to work. Google has that. Most
companies does not.
In article <memo.20251014170713.10624x@jgd.cix.co.uk>,
John Dallman <jgd@cix.co.uk> wrote:
In article <10ckadi$7dr$1@reader2.panix.com>,
cross@spitfire.i.gajendra.net (Dan Cross) wrote:
I never quite got the business play behind Java from Sun's
perspective. It seemed to explode in popularity overnight, but
they never quite figured out how to monetize it; I remember
hearing from some Sun folks that they wanted to set standards
and be at the center of the ecosystem, but were content to let
other players actually build the production infrastructure.
The trick with monetising something like that is to price it so that
customers find it far cheaper to pay than to write their own. However,
you still need to be able to make money on it. I've seen this done with a
sliding royalty scale.
However, this kind of scheme definitely would have clashed with the
desire Sun had to make Java a standard piece of client software. It may
have been doomed to unprofitability by the enthusiasm of its creators.
I think that's a really insightful way to put it.
My sense was that they overplayed their hand, and did so
prematurely relative to the actual value they were holding onto.
I mentioned Microsoft and Java on the client side: I believe
that they were largely responsible for failure of Java desktop
applications (and the supporting ecosystem) to take root. As I
recall, at the time, MSFT tried to license Java from Sun: Sun
said no, and I'm quite sure that McNealy was positively giddy
about it as well. However, I think in doing so, Sun gravely
underestimated Gates-era MSFT, because then Microsoft very
publicly said, "we're going to wait and see whether the industry
adopts Java on the desktop." But, since Microsoft was the
biggest player in that space, the rest of the industy waited to
see what Microsoft would do and whether they would support it on
Windows: the result was that Java no one adopted it, and so it
never saw widespread client-side adoption.
Oh sure, it had some
adoption in mobile phone type applications, but util Android
(which tried to skirt the licensing issues with Dalvik) that
was pretty limited.
Anyway, while Microsoft stalled, they did
C# in the background, and when it was ready, they no longer had
any real need for Java on the client side.
The framing that the web rendered Java on desktops obsolete is
incomplete. Certainly, that was true for _many_ applications,
as the web rendered much of the client-side ecosystem obsolete,
but consider things in Microsoft's portfolio like Word, Except,
PowerPoint, and so on. Those remained solidly desktop focused
until 360;
one never saw credible competitors to that in Java,
which was something Sun very much wanted (recall McNealy's
writing at this time about a "new" style of development based
around open source and Java).
Similarly, investment in C# shows
that they weren't quite ready to move everything to the web;
On 10/15/2025 7:01 PM, Lawrence DrCOOliveiro wrote:
On Wed, 15 Oct 2025 15:33:20 -0500, Craig A. Berry wrote:
I believe Arne's point was the fairly obvious one that a retail chain
or a hospital chain does not need to and cannot afford to maintain,
for example, their own operating system.
Do you think that is hard to do?
Hire enough experts to have people that know the code base of every
critical part: Linux kernel, glibc etc. probably 50-100 million lines of code: bloody expensive. We are talking hundreds engineers - and not just
any engineers but top engineers.
In article <10cpc9g$191j$2@dont-email.me>,
Arne Vajh|+j <arne@vajhoej.dk> wrote:
And one facet of GCP is that Google is taking
over OS support from Redhat/Canonical/SUSE when
companies moves their workload from on-prem to
GCP managed services. Linux support is their
business.
Do you mean ContainerOS? That's just a distro.
On 10/15/25 7:16 AM, Dan Cross wrote:
In article <10cmovf$3a740$1@dont-email.me>,
Arne Vajh|+j <arne@vajhoej.dk> wrote:
On 10/13/2025 10:03 PM, Lawrence DrCOOliveiro wrote:
On Mon, 13 Oct 2025 21:20:43 -0400, Arne Vajh|+j wrote:
On 10/13/2025 8:20 PM, Lawrence DrCOOliveiro wrote:
On Mon, 13 Oct 2025 19:26:56 -0400, Arne Vajh|+j wrote:
Enterprises with a need to document support can not just hire a
random consultant when the need arrive.
If something is mission-critical and core to their entire business, >>>>>> they want a staff they can rely on, completely, to manage that
properly.
Few/no CIO's want to support the hundreds of millions of lines
of open source code their business rely on themselves.
The whole point of having all that code is that they didnrCOt need to write
it themselves.
Yes. But they want free beer more than free speech.
You have to take responsibility for your own business, donrCOt you?
They don't want to write or maintain their own OS.
They don't want to write or maintain their own platform
software (web/app servers, database servers, message queue
servers, cache servers etc.).
They don't want to write or maintain their own tools
(compilers, build tools, IDE's, source control, unit
test frameworks etc.).
None of that stuff is their business.
They want to focus on their business the applications
that help them produce and sell whatever products
or services.
Every single one of the FAANG companies do all of those things.
In other words, hardly anyone.
At Google, we used to joke that, "not only does Google reinvent
the wheel, we vulcanize the rubber for the tires." Spanner,
Piper/Fig/Jujutsu, Prodkernel/ChromeOS/Android, CitC, gunit, Go
(not to mention the work on LLVM/Clang), Blaze/Bazel/Skylark,
etc, are all examples of the things you mentioned above. And
that's not even to mention all the custom hardware.
For organizations working at hyperscale, there comes a point
where the off-the-shelf solutions simply cannot scale to meet
the load you're putting on them.
At that point, you have no choice but to do it yourself.
You're kinda going in circles here by arguing that very big companies
whose business is to make their own technology need to make their own >technology.
I believe Arne's point was the fairly obvious one that a
retail chain or a hospital chain does not need to and cannot afford to >maintain, for example, their own operating system.
On 10/15/2025 8:26 PM, Dan Cross wrote:
In article <10cpc9g$191j$2@dont-email.me>,
Arne Vajh|+j <arne@vajhoej.dk> wrote:
And one facet of GCP is that Google is taking
over OS support from Redhat/Canonical/SUSE when
companies moves their workload from on-prem to
GCP managed services. Linux support is their
business.
Do you mean ContainerOS? That's just a distro.
I am talking about that like 10 years ago a company
would run like:
their application + their database server
RHEL [paying Redhat for Linux support]
ESXi
on-prem HW
but now they may run as (assuming Google customer):
their application in GKE + database as GCP managed service
whatever Linux Google want to use [paying Google for Linux support as
part of what they pay for the cloud services]
Linux with KVM
Google HW
Amazon, Microsoft and Google are taking revenue away
from Redhat (IBM). They have de facto gotten into
the Linux support business.
On 10/15/2025 7:58 AM, Dan Cross wrote:
[snip]
Oh sure, it had some
adoption in mobile phone type applications, but util Android
(which tried to skirt the licensing issues with Dalvik) that
was pretty limited.
Almost all the 3 millions apps available for the 3 billion
Android phones are written in Java or Kotlin. Not particular limited.
Anyway, while Microsoft stalled, they did
C# in the background, and when it was ready, they no longer had
any real need for Java on the client side.
MS started .NET and C# after they were forced to drop their
Java.
Anders Hejlsberg was actually headhunted from Borland to
do MS Java. And when that was no longer a thing he moved
on to creating .NET and C#.
The framing that the web rendered Java on desktops obsolete is
incomplete. Certainly, that was true for _many_ applications,
as the web rendered much of the client-side ecosystem obsolete,
but consider things in Microsoft's portfolio like Word, Except,
PowerPoint, and so on. Those remained solidly desktop focused
until 360;
What moved to web in the early 00's were all the internal
business app frontends. The stuff that used to be done on
VB6, Delphi, Jyacc etc..
Mostly trivial stuff but millions of applications requiring
millions of developers.
MS Office and other MSVC++ MFC apps may have been difficult to
port to web at the time, but it would also have been difficult
to come up with a business case for it - that first showed up
when MS had a cloud and could charge customer per user per month
for it.
one never saw credible competitors to that in Java,
which was something Sun very much wanted (recall McNealy's
writing at this time about a "new" style of development based
around open source and Java).
OpenOffice owned by Sun at the time actually did implement
some stuff in Java.
But neither as OpenOffice as office package nor Java as language
for desktop apps ever took off.
Similarly, investment in C# shows
that they weren't quite ready to move everything to the web;
????
One of the main areas for C# is web applications ASP.NET and
was so from day 1.
(not everybody may like ASP.NET web forms, but that is
another discussion)
In article <10cpeu9$26ht$1@dont-email.me>,^^^ ^^^^^^^^^^^^^^^>>
Arne Vajh|+j <arne@vajhoej.dk> wrote:
On 10/15/2025 8:26 PM, Dan Cross wrote:
In article <10cpc9g$191j$2@dont-email.me>,
Arne Vajh|+j <arne@vajhoej.dk> wrote:
And one facet of GCP is that Google is taking
over OS support from Redhat/Canonical/SUSE when
companies moves their workload from on-prem to
GCP managed services. Linux support is their
business.
Do you mean ContainerOS? That's just a distro.
I am talking about that like 10 years ago a company
would run like:
their application + their database server
RHEL [paying Redhat for Linux support]
ESXi
on-prem HW
but now they may run as (assuming Google customer):
their application in GKE + database as GCP managed service
part of what they pay for the cloud services]
Linux with KVM
Google HW
Not quite how the stack is structured.
Amazon, Microsoft and Google are taking revenue away
from Redhat (IBM). They have de facto gotten into
the Linux support business.
Not really. They're taking revenue away from Broadcom/VMWare,
perhaps, and probably from Dell, HPE, and Lenovo. But if you
want to run RHEL on a VM on Google's cloud, they won't stop you. https://cloud.google.com/compute/docs/images/os-details
In article <10cpebq$26b5$1@dont-email.me>,
Arne Vajh|+j <arne@vajhoej.dk> wrote:
On 10/15/2025 7:58 AM, Dan Cross wrote:
[snip]
Oh sure, it had some
adoption in mobile phone type applications, but util Android
(which tried to skirt the licensing issues with Dalvik) that
was pretty limited.
Almost all the 3 millions apps available for the 3 billion
Android phones are written in Java or Kotlin. Not particular limited.
...but not running on the JVM or using the JRE.
On 10/15/2025 9:01 PM, Dan Cross wrote:
In article <10cpebq$26b5$1@dont-email.me>,
Arne Vajh|+j <arne@vajhoej.dk> wrote:
On 10/15/2025 7:58 AM, Dan Cross wrote:
[snip]
Oh sure, it had some
adoption in mobile phone type applications, but util Android
(which tried to skirt the licensing issues with Dalvik) that
was pretty limited.
Almost all the 3 millions apps available for the 3 billion
Android phones are written in Java or Kotlin. Not particular limited.
...but not running on the JVM or using the JRE.
True.
But the difference is not that big.
[snip]
On 10/15/2025 8:51 PM, Dan Cross wrote:^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
In article <10cpeu9$26ht$1@dont-email.me>,^^^ ^^^^^^^^^^^^^^^>>
Arne Vajh|+j <arne@vajhoej.dk> wrote:
On 10/15/2025 8:26 PM, Dan Cross wrote:
In article <10cpc9g$191j$2@dont-email.me>,
Arne Vajh|+j <arne@vajhoej.dk> wrote:
And one facet of GCP is that Google is taking
over OS support from Redhat/Canonical/SUSE when
companies moves their workload from on-prem to
GCP managed services. Linux support is their
business.
Do you mean ContainerOS? That's just a distro.
I am talking about that like 10 years ago a company
would run like:
their application + their database server
RHEL [paying Redhat for Linux support]
ESXi
on-prem HW
but now they may run as (assuming Google customer):
their application in GKE + database as GCP managed service
whatever Linux Google want to use [paying Google for Linux support as
part of what they pay for the cloud services]
Linux with KVM
Google HW
Not quite how the stack is structured.
Amazon, Microsoft and Google are taking revenue away
from Redhat (IBM). They have de facto gotten into
the Linux support business.
Not really. They're taking revenue away from Broadcom/VMWare,
perhaps, and probably from Dell, HPE, and Lenovo. But if you
want to run RHEL on a VM on Google's cloud, they won't stop you.
https://cloud.google.com/compute/docs/images/os-details
If someone has a strong desire to do cloud like they did 10
years ago, then buying GCE instances, installing RHEL,
installing OpenShift, installing database, installing
application and manage everything is certainly still an option.
But I was very explicit above talking about managed services.
Managed Kubernetes and managed database. GKE not GCE.
Again I wonder if you read what you are replying to.
I've said this before in this group, but the homogeneity of
modern computing does not strike me as a universally good thing.
There are economies of scale one can leverage, to be sure, but
just as monocultures aren't robust against external threats in
biological systems, I can't help but think that the same is true
of computing systems.
It felt like there was a time when we had built hetergeneous
systems that were at least reasonable to manage; these days, I
think we'd know how to do much better. But the diversity of
systems and platforms common 30 years ago are mostly gone, and
we're left with essentially three buckets: Windows, Linux, and
a small sliver of "everything else". Not great.
In article <10co28b$isl$1@reader2.panix.com>,
cross@spitfire.i.gajendra.net (Dan Cross) wrote:
I've said this before in this group, but the homogeneity of
modern computing does not strike me as a universally good thing.
There are economies of scale one can leverage, to be sure, but
just as monocultures aren't robust against external threats in
biological systems, I can't help but think that the same is true
of computing systems.
Yup. I much enjoyed freaking out my old boss the day the ILOVEYOU e-mail >virus hit. Lots of people at work were reading their mail on Windows and
got hit. I was reading mine using Netscape on an HP-UX PA-RISC box and
read the Visual Basic virus code with some interest. Some explaining that >neither HP-UX nor Netscape could run this code, so I was safe, was
necessary.
Some years later, when I had a new boss, the old one spotted me reading a >book about how buffer overflows and other security holes actually work.
He was concerned, and felt that staff should not know these things. My
new boss gave me a meta-instruction: if doing something reasonable
worries the old boss, keep on doing it.
It felt like there was a time when we had built hetergeneous
systems that were at least reasonable to manage; these days, I
think we'd know how to do much better. But the diversity of
systems and platforms common 30 years ago are mostly gone, and
we're left with essentially three buckets: Windows, Linux, and
a small sliver of "everything else". Not great.
The heterogeneity shows up at different levels these days. VMware tried
to enforce a monopoly, and now lots of different virtualisation systems
are getting more popular. Tintri is taking market share from NetApp, and
so on.
| Sysop: | Amessyroom |
|---|---|
| Location: | Fayetteville, NC |
| Users: | 54 |
| Nodes: | 6 (0 / 6) |
| Uptime: | 03:35:00 |
| Calls: | 743 |
| Files: | 1,218 |
| Messages: | 188,321 |