Every day people make millions of successful temperature measurements. The people who make these measurements frequently identify ways in which they would like temperature measurement to be improved. Typically people would like results with lower uncertainty and at lower cost, particularly at extremes of temperature. But nobody has ever expressed — to me at least — any dissatisfaction with the definition of what we mean by “one degree”. Nonetheless, in 2018, the CIPM will introduce a new definition of the kelvin and degree Celsius. In this paper I discuss the background to this redefinition and explain why — despite not addressing users' pressing needs — the redefinition is desirable and indeed long overdue.