back to article Fujitsu confirms end date for mainframe and Unix systems

Fujitsu has confirmed the end of the road for its mainframes and Unix server systems. It will cease to sell both by the end of this decade, with support services continuing for a further five years. Customers are by then expected to have migrated to the cloud. The tech giant's plans were revealed in a notice posted to the …

  1. Anonymous Coward
    Anonymous Coward

    How do you think I'm going to get along, without you when you're gone.

    You took me for everything that I had and kicked me out on my own

    Applies pretty well here.

    1. teknopaul

      A decent heads up is acceptable

      Probably just fine for a good long while, we have old Vax systems running.

      Private companies have to look after the bottom line or they go out of business and all customers are affected.

    2. kurkosdr

      At the end of the day, all proprietary software has an EOL date.

      Maybe there is something to be said here about relying on a proprietary software stack (that the vendor can EOL anytime) for your business needs.

      In the consumer sector, Apple can sell customers a Power Mac in spring of 2005, announce a switch to Intel during the summer of that same year, and then proceed to completely screw PowerPC users over a mere 4 years later with Snow Leopard, and it's just how things are done in the consumer sector (current Intel Mac users are about to get the same treatment). Similarly, Nvidia can sell RTX 20 series graphics cards with 3D Vision support to customers interested in stereoscopy in 2018 and then proceed to completely screw them over in 2021 by EOL'ing 3D Vision from new drivers. Again, that's all expected in the consumer sector.

      But when your business relies on a software stack and "transitions" cost money and man-hours, you can't have any of that.

      And this also goes to you fancy people in "smart casual" jeans building apps on "serverless" / "app engine" clouds that are near-impossible to port to other "serverless" / "app engine" clouds, not just suits buying mainframes from Fujitsu. Long story short, don't do it unless you absolutely have to.

      Have you tried using docker containers managed by Kubernetes?

  2. Pascal Monett Silver badge
    WTF?

    Support for five more years ?

    It's a mainframe, not a laptop. You support a mainframe for 25 years, not five.

    If Fujitsu thinks that it's going to sell a mainframe that only has five years of support left, I think Fujitsu is in for a bad surprise.

    They may as well stop making them now - as long as they still offer support until 2035, they just might still sell one or two.

    1. Snake Silver badge

      Re: Support for five more years ?

      "If Fujitsu thinks that it's going to sell a mainframe that only has five years of support left, I think Fujitsu is in for a bad surprise."

      "If Fujitsu thinks that it's going to migrate its clients to their private cloud, rather than said clients migrating to far more experienced and integrated suppliers like AWS, I think Fujitsu is in for a bad surprise."

      FIFY

      1. John Brown (no body) Silver badge

        Re: Support for five more years ?

        I guess it depends, as per the article, on how tied their software is to the Fujitsu mainframe architecture. Anyone able to move to a different cloud is probably also able to move to different hardware and not move to any cloud.

      2. Admiral Grace Hopper

        Re: Support for five more years ?

        If Fujitsu thinks that it's going to migrate its clients to their private cloud

        Fujitsu’s private cloud - K5 - died on its arse years ago. Their cloud solution is reselling Azure.

    2. PM from Hell

      Re: Support for five more years ?

      When I was involved in mainframe support the replacement cycle was 3-5 years to keep increasing capacity. There will be some customers who will need to increase capacity to either keep decent levels of system performance or to provide additional capacity to support the migration work. Whilst in some cases this can be provided by an additional lower spec mainframe some will need the additional power on a single mainframe.

    3. itsborken

      Re: Support for five more years ?

      Alpha and Itanium come to mind.

    4. Justthefacts Silver badge

      Re: Support for five more years ?

      “Your Crisis is not my emergency”

      Fujitsu evaluated the business case for *developing new mainframe chips*, and decided there wasn’t one. Back in 2017, when they decided to stop after SPARC64 XII. Five years later, it’s five years later. I’m not sure how this comes as a surprise to anybody.

      IBM Power 10 is out now (on Samsung 7nm) to power the IBM z16 this year. A chip that complex probably cost IBM 1-2billion to develop. There is a roadmap for POWER11 in 2026, which would likely go on 3nm. If they strip the development absolutely to the bone, and just go for a pure die-shrink (zero new stuff) and packaged chiplets, they can probably keep the development costs under 5billion. But if there were to be a POWER12 on 1nm…..there’s just no way that is costing less than 20billion. For sales of only 2.5bn a year. POWER12 is just never going to happen, and I suspect that POWER11 might not either if IBM had stronger management.

      Mainframes might be great for your business, but the development costs are ballooning beyond any credible ongoing business case.

      1. seven of five
        Trollface

        Re: Support for five more years ?

        Wow, IBM would sure benefit from just the facts of your analysis.Try to sell it to them.

        1. Justthefacts Silver badge

          Re: Support for five more years ?

          They are sold, for many years now.You just haven’t been paying attention. They are mostly a software company now - just not a very good one.

      2. Hurn

        Re: Support for five more years ?

        Excuse me, but the POWER family of chips does not power Z.

        The requirements for POWER 3 through Power 10 are much different from the Z Mainframe CPUs, although it is true both CPUs have benefited from process improvements.

        One example:

        A P10 chip, running at, let's say 300 Watts, is able to be cooled using forced air (fans) with big heatsink.

        Z systems use liquid cooling (with radiator and anti-freeze/water mixture), because their CPU modules run at much higher power levels, and are more densely packed.

        1. Justthefacts Silver badge

          Re: Support for five more years ?

          I apologise, you are right….

          But the point still remains: IBM mainframe revenue is declining, not even stable. In 2017 it was just $1.6billion. *As a business*, rather than a hobby, you can’t allocate more than $300m per year for R&D. That just doesn’t pay for a leading-edge chip design in the modern world, and design costs are doubling every 3-4yrs.

      3. Anonymous Coward
        Anonymous Coward

        Re: Support for five more years ?

        Do you always pull numbers out of your ass, or do they sometimes appear to you in visions as well?

        Follow-up question: do you work for Wired magazine?

        1. Justthefacts Silver badge

          Re: Support for five more years ?

          Data?

          https://www.mckinsey.com/industries/advanced-electronics/our-insights/semiconductor-design-and-manufacturing-achieving-leading-edge-capabilities

          They estimate that design costs for 5nm chip are $540m. But that’s *industry average for a SoC*, which is mostly just a process shrink of the previous mid-range smartphone chip. The most complex mainframe chip on the planet? Hell yes that’s three or four times the cost. And you can perfectly see that the cost is nearly doubling per generation. Intel R&D is $15bn *per year*, of which half is chip design. How many really different chips (as opposed to variants) is that? Five? So, no, I don’t work for IBM. But you think they are really getting away with spending less than a couple of billion for a 3nm mainframe chip? I’ll have what you’re smoking.

          Shorter: I used to manage complex chip developments for a living, although nothing on that scale. You clearly don’t.

    5. ArrZarr Silver badge
      Devil

      Re: Support for five more years ?

      Now admittedly, I'm bumbling into this conversation from a standing start, but I'd guess it's likely that they don't have a big production line for these, especially considering the low volume being described by the article.

      You're probably right that they shouldn't expect to sell many more of the mainframes, but they aren't going to throw away the money from being able to put a couple more together on demand, are they?

      Beyond that, they've been clear about timelines until support ends so anybody buying one will have the information readily available (and any subsequent failures due to lack of support post-2035 are on the customer for making a poor decision).

      Icon: Devil's advocate

      1. Roland6 Silver badge

        Re: Support for five more years ?

        >but I'd guess it's likely that they don't have a big production line for these

        When I looked in detail at the mainframe sector back in the late 1990's, IBM Z-series was the leader but it only had circa 35,000 customers worldwide. Most of the others only had a couple of thousand if they were lucky. My recommendation to the board was to "rightsize" their own mainframe-base (under 2000 customers) and to either transition customers to their datacenter grade Unix servers or if they really needed a mainframe - Z-series... In both cases the big carrot was the getting customer data off batch/tape into relational/data warehouse and so improve customer service etc.

        I suspect things haven't changed very much and Fujitsu are probably having to come to terms with the cost of servicing only a few thousand (mainframe) customers, where they can't charge supercomputer prices and support costs.

    6. teknopaul

      Re: Support for five more years ?

      I think it's decent to sell new ones for five years, people may way to buy spares for existing systems knowing they will not be available in the future.

      Clearly not a good idea to base a new project on this hardware.

    7. Anonymous Coward
      Anonymous Coward

      Re: Support for five more years ?

      Totally agree but it's getting people to run them that's getting harder. Last mainframe I worked on was back in 2003 just as it being replaced by a couple of mini systems with same power. The incumbant main op/admin was retiring but he agreed to stay for 2 years to see the transition through for extra bonuses.

      I remember contracting at a service company once, a group of old boys used to sit in the corner of the room, they hardly ever appeard to do any work. They were paid simply because they knew a specific old mainframe system that a customer had. They spent all day reading the paper but the customer was terrified the system would fail them that they paid huge sums of money for these old boys to babysit their precious system while they were working on plans to replace it.

      It's not the system, it's finding people who know how to manage it that's getting harder and more expensive. Who the hell wants to work in IT past 55?! No one, I've no intention of working beyond 60 thanks very much, I couldn't care less even I'm last person around who knows something, when i can retire I'm off mate!

  3. heyrick Silver badge

    "as migrating from a mainframe is not a simple operation"

    Might be a dumb question - but is there no such thing as a mainframe emulator that can run on more modern hardware?

    1. Anonymous Coward
      Anonymous Coward

      You can emulate a mainframe's instruction set and OS but how are you going to emulate the hardware optimized for transaction processing, I/O, and 99.999+% uptime?

      1. Anonymous Coward
        Anonymous Coward

        Relax! Amstrad's making a comeback and plans to relaunch the Sinclair XL.

      2. Anonymous Coward
        Anonymous Coward

        This old shtick again?

        "Transaction processing" is a sad joke. Mainframe hardware is SLOWER at doing ANY kind of processing than modern servers. The same goes for I/O. And modern uptime is achieved with redundancy, not one big box.

        Emulation is fine. The only people who buy mainframes in the 21st century are idiots stuck in the past.

        1. x-IBMer

          All you've shown here is how little you know about mainframe hardware and transaction processing.

        2. Anonymous Coward
          Anonymous Coward

          Absolutely - just look at Sabre compared to Amadeus and Travelport

          Of the 3 main GDS (global distribution systems - which manage huge databases for airline fares, hotel, hire car availability, etc.) Sabre is the only one not running on a mainframe AFAIK. They migrated to RH and commodity hardware some time ago. They’re doing very well thank you. Amadeus still run the majority of airline/port systems on legacy mainframe hardware.

          It’s perfectly possible to migrate if you’ve got the money, vision and ability to deliver.

          1. Paul Crawford Silver badge

            Re: Absolutely - just look at Sabre compared to Amadeus and Travelport

            It’s perfectly possible to migrate if you’ve got the money, vision and ability to deliver

            True, but how many large institutes have even one of them?

          2. Anonymous Coward
            Anonymous Coward

            Re: Absolutely - just look at Sabre compared to Amadeus and Travelport

            "Amadeus still run the majority of airline/port systems on legacy mainframe hardware"

            Always wondered why the UI for that sucked... Good for finding actual flights though.

          3. Anonymous Coward
            Anonymous Coward

            Re: Absolutely - just look at Sabre compared to Amadeus and Travelport

            https://amadeus.com/en/insights/blog/celebrating-one-year-fully-open-systems

            Celebrating one year of running fully on open systems Slightly more than a year ago, towards the end of 2018, Amadeus shutdown its last mainframe; this milestone came on the heels of an even bigger step completed in June 2017, when we retired our entire TPF mainframes complex.

            1. Roland6 Silver badge

              Re: Absolutely - just look at Sabre compared to Amadeus and Travelport

              >" Over time, we found that the mainframe became slow and rigid to take on the challenges of the digital age. It can take engineers three to four months just to train on them, and new graduates have rarely been motivated to learn such out-dated technology."

              I wonder what would happen if there was an open-source mainframe OS...

              Perhaps given Fujitsu are end gaming their mainframes, they could put a toe in the water and make their mainframe OS's open-source; given their age its not like they would be infringing any current patents...

              Part of the problem mainframes had was that Universities didn't have the funding to buy them and thus expose undergraduates to them, hence why so many come out of University with limited knowledge of the range of computer systems being used in the world outside of University computer departments.

            2. Anonymous Coward
              Anonymous Coward

              Re: Absolutely - just look at Sabre compared to Amadeus and Travelport

              Congratulations on making your public facing systems look like they’re still in the 1980s ;)

              1. philstubbington

                Re: Absolutely - just look at Sabre compared to Amadeus and Travelport

                … and this just in……

                https://techcrunch.com/2022/04/07/what-startups-can-learn-from-amadeus-migration-to-cloud/

    2. Anonymous Coward
      Anonymous Coward

      is there no such thing as a mainframe emulator that can run on more modern hardware?

      The problem is finding modern hardware that can provide the same performance as a modern mainframe. Maybe a large multi-system, multi-processor cluster could handle the CPU load, but perhaps not the 20,000+ connected terminals.

      1. Anonymous Coward
        Anonymous Coward

        Ooooo, 20k terminals! What a HUGE number!

        Yes, that's sarcasm. Any moderately busy website has 100x that number of concurrent users, and doesn't run on a mainframe because NO mainframe could handle it.

        20k terminals needs about the processing and I/O power of your average Raspberry Pi.

        1. martinusher Silver badge

          >Any moderately busy website has 100x that number of concurrent users, and doesn't run on a mainframe because NO mainframe could handle it.

          The legacy mainframe is probably no more powerful than a cluster of server blades. The rationale for keeping it is "if it works leave it alone" -- the cost, dislocation and risk needed to replace it may not be justified. There's also something to be said for keeping business data under local control so the mindset of 'just move it to the cloud' doesn't necessarily work unless you happen to own that cloud.

          Websites are often buggy and are surprisingly labour intensive -- their code needs constant attention with updates, bug fixes and protection against intrusion. What happened to British Airways yesterday isn't an isolated case.

        2. Anonymous Coward
          Anonymous Coward

          Hmm, someone who doesn't understand the difference between a process and a near-realtime driver.

          How well would your RPi cope with 20K devices plugged into a USB bus, each connected to 2-10 processes?

    3. thames

      There are emulators for IBM mainframes that run on Linux. I don't see why the same couldn't be done for Fujitsu systems.

      The problem will be the operating system and other associated software. At the last that I heard customers attempting to run the above mentioned IBM emulators run into problems because IBM won't license their OS for use on emulators. I don't know if Fujitsu will take a different position.

      The first paying software development job that I had was reverse engineering an IBM mainframe application to run on a PC. That is, here's the inputs, here's the outputs we desire, write a program which does this on a PC running MS-DOS. I did it for a small fraction of the cost of the proposed charge to be added as a customer on the existing mainframe application, which was running in a EDS data centre (the original "cloud"). I know someone else for whom one of his first paying software development jobs was also to move a mainframe application to a PC. In his case the mainframe was nearing end of life and moving to another mainframe was seen as not really providing a long term solution.

      There were obviously small, specialized applications, but loads of these were around at the time as big companies had IT departments that preferred doing things on mainframes which smaller companies used PCs for. Most mainframe applications didn't involve anything that couldn't be done with a fast enough PC with a big enough hard drive provided you knew what you were doing and weren't trying to just duplicate how the mainframe application approached the problem.

      There's probably no single answer to this sort of question. Each company will have to look at their own individual situation and figure out how to deal with it. The biggest problems for most companies probably aren't going to be the technical ones, bur rather the ones involving their own internal bureaucracies which will attempt to throw barriers in the way of any sort of change.

    4. NorthIowan

      Re: mainframe emulator

      Unisys went all emulated back in 2012.

    5. Admiral Grace Hopper

      That’s precisely what happened with ICL’s VME mainframe OS which runs under RHEL. VME is owned by Fujitsu as well.

    6. An_Old_Dog Silver badge

      Remember the data transfer

      A significant problem will be copying all those datasets from IBM DASD devices to the emulator host's (or hosts') hard drives, in a timely manner.

  4. adam 40 Silver badge

    Oh dear what will the Post Office do now?

    Where will they find another supplier willing to perjure itself?

    1. Anonymous Coward
      Anonymous Coward

      Re: Oh dear what will the Post Office do now?

      Somewhere over the Horizon, perhaps?

    2. Doctor Syntax Silver badge

      Re: Oh dear what will the Post Office do now?

      Once it's in the cloud things will be so much more obfuscated they won't need actual perjury.

    3. teknopaul

      Re: Oh dear what will the Post Office do now?

      "Where will they find another supplier willing to perjure itself?"

      Let's make them a list.

  5. spold Silver badge

    General trend?

    Even IBM seems to be touting IBM Z in the cloud more than IBM Z on-prem, probably because it turns it into recurring revenue rather than a one-time sale - though it seems they may have the mainframe market to themselves soon.

    1. Justthefacts Silver badge

      Re: General trend?

      As elsewhere….IBM have just released their new POWER10 chip for z16.

      If they make a POWER11, it will no more than break even. But a POWER12 would break the company. So, they are very well aware their mainframe business has to finish within about 15 years, apart from maintenance. How much do they want to invest, vs pivot. But they don’t seem to have any idea what they want to pivot *to*. Watson was just stupidity.

      All the obvious commercially sensible pivots could have been executed 10-20 years ago, but now the time has gone.

  6. Anonymous Coward
    Anonymous Coward

    Should think a lot of UK Government systems will be affected

    When Fujitsu took over ICL they inherited a shed load of Government systems.

    I'd think many of these Ministries would prefer not to host these on Cloud based systems, private cloud or otherwise.

    1. Doctor Syntax Silver badge

      Re: Should think a lot of UK Government systems will be affected

      I think you overestimate HMGs preference to maintaining security versus employing sufficient people with degrees in non-U subjects such as engineering.

    2. Anonymous Coward
      Anonymous Coward

      Re: Should think a lot of UK Government systems will be affected

      Many of these have migrated off VME, the rest are in the process of being migrated. Fujitsu have tried to push VME users to an as-a-service model, government have more often than not pushed back.

      In other news, there is a large amount of government systems already hosted on Azure and AWs. Many other cloud providers were tried, many of which are no longer with us, but those were the eventual winners.

    3. innominatus

      Re: Should think a lot of UK Government systems will be affected

      Bring back ICL? Or LEO!

      1. Dave559 Silver badge

        Re: Should think a lot of UK Government systems will be affected

        I can foresee that the line, "Well, yes, Minister, it certainly will involve as much tea and coffee as you desire, and plenty of delicious cakes and waitresses", could definitely assist with the budgetary approval for such a project… (and, yes, "delicious" applies to both of the following nouns)

      2. spold Silver badge

        Re: Should think a lot of UK Government systems will be affected

        Fujitsu purchased ICL

  7. ecarlseen

    The Cloud does not replace a mainframe.

    As mentioned before, organizations use mainframes for applications that cannot ever go down, ever, for any reason. Mainframes usually deliver somewhere close to their 99.999% uptime guarantees in the real, messy world that we actually live in. It's far from unheard of for them to have zero downtime over the course a year. How many other systems offer that kind of uptime anywhere outside of their marketing materials?

    I mean, there's in article in El Reg today about how many vendors lie through their teeth about their uptime metrics, and everyone with experience is nodding along as they read.

    https://www.theregister.com/2022/02/24/cloud_service_status_pages_fail/

    In fairness, not many applications really require five 9s of availability, but for those that do a mainframe is still a very respectable option.

    1. rcxb1

      Re: The Cloud does not replace a mainframe.

      > It's far from unheard of for them to have zero downtime over the course a year. How many other systems offer that kind of uptime anywhere outside of their marketing materials?

      I've maintained lots of Linux servers with multiple years of uninterrupted uptime. Or main order entry system was up over 4 years with hundreds of sessions at a time at peak, and operating around the clock, before a scheduled power outage required ending its run.

    2. Anonymous Coward
      Anonymous Coward

      Re: The Cloud does not replace a mainframe.

      How many other systems offer that kind of uptime anywhere outside of their marketing materials?

      Well here's one of the DNS/mail/web/imap servers I manage:

      % uptime

      11:42am up 2294 days, 7:38, 1 user, load averages: 0.47, 0.54, 0.51

      No marketing materials were involved. It runs BSD.

      1. Anonymous Coward
        Anonymous Coward

        Re: The Cloud does not replace a mainframe.

        Not installed any kernel updates then, or done the needed reboot if you have.

        1. Anonymous Coward
          Anonymous Coward

          Re: The Cloud does not replace a mainframe.

          There is nothing in the newer kernels that is needed. If there was, the kernel would have been upgraded.

          In short, it it ain't broke don't fix it!

          1. kurkosdr

            Re: The Cloud does not replace a mainframe.

            You are aware that security vulnerabilities mean a kernel is badly, badly broken, right?

    3. Anonymous Coward
      Anonymous Coward

      bollocks to five 9s!

      "In fairness, not many applications really require five 9s of availability, but for those that do a mainframe is still a very respectable option."

      In a previous job at a very large company, the email team proudly announced they'd achieved 5 nine's availability. They weren't pleased when I asked them to choose the 30 seconds in the current year for DNS service to be switched off so it could match that result.

    4. PriorKnowledge
      Happy

      Windows does just fine

      As long as you keep it off the internet, never patch it, cluster the roles, stick the cluster endpoint behind multiple load-balanced reverse proxies and very strictly limit allowed inputs/queries.

      1. Anonymous Coward
        Anonymous Coward

        Re: Windows does just fine

        Windows servers don´t last that long older ones got problems after 230 days. It was a fujitsu siemens one

        1. Plest Silver badge
          Facepalm

          Re: Windows does just fine

          Is it Win2008R2 that has a bug that causes it to fail after 450 days of uptime? We had a couple that were about to clock over and people started panicking. Personally I reboot all Windows boxes on schedules at various times once a month, forces patches, flushes, etc.

          Modern kit and O/S are designed to kicked over every so often. Uptime is like sitting still too long with a bad back, the second you move you'll break something serious and worse you have no choice but to do it no matter how painful it's going to be. Bad back, you keep moving! Reboot on schedule when it's quiet and flush the shite!

    5. Dunstan Vavasour

      Reliable Apps vs Reliable Systems

      System reliability matters when you have monolithic applications. Once you build the resilience into the application layer, system reliability doesn't need to have as many nines after the point.

      Now, that's not to say that distributed apps with distributed data is a silver bullet - you still need to do grown up failure mode analysis - but mainframes aren't your only path to all those nines.

  8. John Savard

    Possible Futures

    If I remember correctly, Fujitsu's mainframes are compatible with the 32-bit version of IBM's 370 architecture. So that is a possible path of migration.

    As for getting customers to migrate to Fujitsu's cloud - well, if the mainframe applications are only licensed to run on Hercules if they're run on Hercules in Fujitsu's cloud, problem solved.

  9. Bitsminer Silver badge

    2034

    Support services for both portfolios will continue for five years afterwards, meaning these will end in 2034 for Unix servers.

    Just in time to avoid the Unix/Linux 2038 bug.

    https://en.wikipedia.org/wiki/Year_2038_problem

  10. Henry Wertz 1 Gold badge

    Cloud hardware

    So, is Fujitsu just suggesting people move from mainframe to an unrelated cloud service? Or is it going to effectively be a mainframe hosted on their end? If so, are they making new mainframe hardware for their own use, or are they just going to keep running the same systems "in the cloud" for some indefinite length of time?

    1. Justthefacts Silver badge

      Re: Cloud hardware

      Emulated OS on cloud commodity server. What’s the problem?

      You were prepared to pay mainframe prices, which easily pays for fully triplicated & redundancy-voted service (not just prime & redundant failover), leaving mucho $ for some nice trips to the Caribbean.

      It’s a perfectly good technical solution. You will be surprised how quickly “received wisdom” about reliability and availability changes once the old mainframe manufacturers no longer have an interest in selling them.

      Once you accept emulated OS, all a mainframe actually consists of is: a proprietary software checkpointing and availability layer, some hardware redundancy on the backplane, a few special instructions, and a particular mix of I/O.

      #1 remains in place. #2 can be disaggregated as a special unit held in Fujitsu data centre. #3 & #4 is mostly FUD anyway and to the extent it’s not can be easily dealt with by massively over-provisioning still at less than half the cost.

      The only thing really special about mainframes is their single-threaded speed…..which is essentially useless. The application is either legacy (in which case 0.1GHz is fine, because it was fine twenty years ago), or modern (in which case it is multi-threaded and parallel).

  11. Roger Kynaston
    Pint

    Fujitsu ICL

    During my stint at a London Borough they had an ICL mainframe. That wasn't my problem but I did have a couple of horrible Team Servers. Come the time when the application was finally replaced I caused a bit of a stir in CAB when I asked to throw them off the roof as part of the decommission process.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like